Over the past year I've built and sold hundreds of thousands of dollars worth of AI chat Bots to clients all over the world most of which were built on no code platforms so in this course I'm going to be teaching you absolutely everything that I know about building AI chat bots so that you can have the same sort of success as a freelancer as a business owner by utilizing this amazing new technology over the past year I've Been on hundreds of Consulting calls and had hundreds of conversations with people within my Discord community and I
keep hearing the same question where can I learn what you know unfortunately I've never really had a good answer for them there's no course there's no degree there's no one YouTube channel that has a beginner to expert resource for AI chatbot building until now and so I'm creating this video as a public service really to provide a complete resource For people wanting to learn this new and valuable skill to create a new source of income to help grow their business us or just to come up to speed on the current state of this new and
exciting technology I've spent weeks preparing this content for you all and what you're about to learn is the result of a year of trial and error in my own chatbot development Journey this video is the resource that I wish I had when I started my journey over a year ago as I'm sure you can already tell by the length of this video we have a lot of information to get through today so I'm going to have timestamps as always so feel free to skip around but these are the things we're going to be covering in
today's video firstly we're going to be quickly diving into what the AI chatbot opportunity really means for people like you and I this includes a brief discussion of the macro factors that are at play what sort of Trends we're Writing here the main drivers for this if you're completely new to my channel and aren't aware of the huge opportunity to sell chatbots to businesses I'd definitely recommend you start from the beginning and listen to that because there's quite a lot of background you need to know before you can start to really see the opportunity that's
in front of us then I'm going to be giving you a deep dive on the different kinds of chatbots that are available to build Within the market and the sort of pros and cons of each then we're going to be jumping into an analysis of the best softwares available this is going from prototyping to full chatbot Builders to tools I'm going to be covering everything so that you guys are aware of what exists on the market and you guys can go straight to the best ones rather than spending all the time I've done working with
the bad ones then we're going to be getting our hands dirty by Breaking down the key concepts of chatbot building and sort of AI in general so that you're able to understand and have a foundation of knowledge that you can build your knowledge on chat box on top of and then finally we're going to be taking all of the information and things that we've just learned and put them into practice by building nine different chat Bots live step by step for you guys to follow along at home the chat Bots we're going To be building
range from extremely basic all the way up to extremely Advanced so you're going to get to go step by step and improve your chatbot building skills all the way through and we're going to be teaching it on one of the most powerful bot boarding platforms so you're building a skill set already that is going to be valuable for years to come and then finally I've thrown in a little bit of a bonus which I think is necessary at the end of this video on How you can actually take the chatbot building skills and building the
chat Bots and start selling it to businesses so all of that is going to be covered at the end of the video once you have the skills required to do so now before we get started there are a few things that you're going to need firstly you have to go and get a notebook you can't watch this video without a notebook and a pin there's just no way that you can watch all of this and not take a couple notes Down so go get a notebook or go get a pin go get a coffee I've
got mine you're also going to need an open mind to these new Concepts that are coming in and also there's a lot of sort of mental resistance I get from people sometimes talking about chat Bots and thinking there's no way that you could sell them for that much you're also going to need an open mind to the new things I'm about to teach you but also to to keep an open mindset really that anything that you Want to learn you can learn it there's no sort of I'm a developer I'm not a developer or I'm
a this Runner that have an open mind and be ready and confident in your ability to learn anything you want if you put your mind to it and I assure you that if you're watching this you definitely can do what I'm going to show you in this video and finally what I need you to do right now is pause this video close all of your tabs come back to this video and save it to your watch It down below leave a like And subscribe while you're down there if you want to support me because you
need absolute focus on this video for the next two hours two and a half hours however long this ends up being I'm prepared to sit here for five or six hours to film this for you guys so I'm sure you can sit there for two hours to listen to this whole thing and take notes so if you've got your notebook you've got your pen you've got your coffee open mind and You're also fully focused on this video and this tab only then we're ready to get started now before you spend two hours listening to some
guy on the internet I'm sure you're going to want to know a little bit of background about me so you know you're not wasting your time and why you should even listen to me at all so a little bit about me my name is Liam otley and I'm an online entrepreneur from New Zealand but now based in Dubai and currently in Europe For the summer I'm the creator of the AI automation agency or Triple A as it is known and I'm also running the largest community of AAA owners in the world we have about 23
000 members in there which I'm teaching to regularly I've been doing online business for nearly five years now and as you can see I've had nearly as many hairstyles as I've had years in the game and through my online businesses I've been fortunate enough to build the life of my dream is where I Live where I want I work with the people I love working with and I get to work on things I really love working on my main focus right now is on building and scaling my AI Automation and AI development company called Morningside
AI as you can see on screen here we have three parts of the company Morningside AI Morningside Automation and our Morningside studio and collectively they're doing about a hundred and twenty thousand dollars per month now so that's Up a little bit from last month if you've seen my previous videos we do a number of different things including building custom AI applications for clients but also complete business automations with AI and AI integration I want to point out that a significant chunk of our monthly revenue is coming from our automation side which is using the exact
same skills that I'm going to be teaching you in this video to build chat Bots using low code and no code Platforms so all of that stuff that you're going to learn is applicable to the exact same kind of stuff that we're using to generate tens of thousands of dollars per month in Revenue so that's a little bit about me about my background my company what I do and why you might want to listen to a two or three hour video from it I'll leave that up to you to decide one last thing before we
get into the video I wanted to let you all know that the waiting list for my AI Automation agency accelerator is now open in my last video I asked you all as my community if you'd be interested in me putting together as some kind of exclusive community that would be led by coaches that are hand-picked by me that are industry experts in their given fields in order to sort of hold your hand and walk you through step by step on your AI automation agency Journey if you've been following me for a while you know that
my method has always been to Give everything away for free and then I can sell on the implementation via Morningside AI my development company however I've been feeling that perhaps this is actually hindering the potential of some of you more than it's actually helping you and it seems like many of you agreed based on the results that we got from the poll that I put out ninety percent of you said that you would be interested if I bought something like this together so I'm going to be Releasing just 150 spots to my exclusive accelerated community
over the coming week the only way to get access to this will be through getting on the waiting list there's going to be a link for you to sign up to it so you can head down below and sign up to the waiting list now if you want to jump on that as soon as possible section number one the AI chatbot opportunity explained in this section we're going to be doing some important groundwork in understanding And and really realizing why this trend is so big why it's such a big opportunity to get into and why
you'd even consider building the skill of building AI chatbot this section is absolutely crucial background for you before you go into the later stages of this video we're essentially establishing the why for you why are you bothering to learn AI chapter development why would you even want to learn how to make these things all of That is going to be absolutely crystal clear by the end of this section well first of all if you're watching this video you probably don't need me to tell you that we are in an absolutely booming Market the AI space
in general but specifically chat Bots as well so we're kind of in a super hot area of a super hot Niche already as you can see on screen the the growth rates of these markets is absolutely exploding this alone is kind of a good enough reason to Be paying attention or at least trying to build skills within the area learning an essential skill within a booming market like chatbot building within AI is really a no-brainer in a lot of cases if you're looking to get into it you're looking to participate and you're looking to make
money now when it comes to applications for chat Bots how they can be applied to provide real world value it's always going to come back to how businesses can use them to improve Themselves with the business to treat their customers better and this is really where the main value of this new technology is going to be evident and for you as someone who's looking to learn this skill as a freelancer to start selling your services or as a business owner to start helping to grow your own business it's important to know what the benefits and
the opportunities are for businesses themselves firstly chatbots open up a new Communication Channel between customers and businesses which can be hugely beneficial you can sell things to them you can get insights from them as we see later here secondly there's a 24 7 availability on these chat bots so you're getting increased customer satisfaction and better service to people globally so you're not just working within maybe the eastern Time or things like this AI chatbots by Nature have sort of cost savings built into them right there they're intended to Automate tasks and replace people by being
artificial intelligences so they're able to make decisions they're able to handle queries without needing to pay them essentially and businesses can actually increase their revenue by using tools like AI chat Bots because it's more of a personalized shopping experience customers can or clients can get access to a personalized service or personalized upsells personalized recommendations that are just not Possible with traditional chatbots or with other marketing methods really and finally in my personal favorite this kind of goes back to our first point of a new customer engagement Channel but the data that you get from the
chatbots and the customer queries and information that's coming through them the transcript information offers a completely new data source to extract insights from an assistant decision making you are going to be getting so Much data from these customers and you're able to analyze that data and make decisions offered on products on sort of strategic moves and things like this this is the really really exciting bit for me personally now with that background on what chatbots are why businesses would even be interested in implementing them it's important to take a bit of a closer look and
really ask why is this happening what is the underlying Trend that's occurring that's Making chatbots become more and more popular that's really carrying this entire thing forward the answer to this question is really the real Revolution that has been started by Ai and I think there's going to be a number of different ones but this is the first one we're experiencing right now and this is the one that we can really focus on and build skills around and make money on now and thanks to Ai and its ability to understand text and voice we are
moving Away from these point-and-click websites towards a world where everything we want to do on computers or with our phone Etc or on websites can be done via either typing it in and telling it what we want or by speaking to it this can be a little bit hard to grasp sometimes so I want to break it down one example of this could be that you wanted to I don't know post on Facebook about you going on holiday right now you'd have to go onto Facebook you'd have to make click click Click you'd have to
type in all the information you'd have to hit enter and you're going to send that and make the post similarly if you wanted to look for a house on the residential website so you could go on and make click click click I want to filter by area by number of bedrooms all of this manual input that you're having to do in order to get your results now in the future that we're moving to towards both of those things can be completely changed the way We do them really so for Facebook you could either type to
your Facebook assistant and say make a post on Facebook about how I'm going holiday soon bank and it can do it all in the background for you or ideally you can do it when Apple gets their things together so sorry through a voice assistant you can just press and hold it and it says hey write me a post on Facebook for me going on holiday in the next week or can you look for houses on this site that Are five bedrooms in this area and it can do it all for you but that's a little
bit far off where we're going with the voice I think that will come soon but in the meantime the chat is the sort of method of input that we're going to right now thanks to the huge success of things like chat gbt people are more familiar than ever with telling a computer or instructing an AI what to do through text now the only question to ask here really is why is this happening Why would consumers start to expect or even demand natural language input with their businesses they interact with rather than point-and-click websites which have
proven to work and we've been using them for decades the answer to this of course is that the human brain by nature is a lazy organ our brains have developed to do the bare minimum in order to survive to conserve energy mental energy physical energy at all costs in order to be as survivable as Possible right so in our modern context let's take some kind of weird shapes but in this case it is the fact of having to think about what you need to do on a website having to go in and physically click and
do all these things that takes more energy than just knowing what you want and if you cut away to communicate to a computer or communicate to a website purely what you need done and you didn't have to do it yourself that would be a huge improvement over point And click to this new sort of natural language input this Revolution and how we interact with computers represents an Outsourcing of the problem solving and of the actual doing from our brains and from our bodies to the computer when we communicate our needs to our computer and tell
it what to do through chat or through natural language all we have to do is think about what we want to do and communicate that and they do the rest you know longer do you have to do the Problem solving of going I need to find a house of this type okay now how do I do that I need to click here and here you just tell it what you want and it does the rest that is a huge benefit and a huge improvement over having to do things manually purely because of the time and
energy required but also the mental strain required to to think about how you want to execute the things you want done and it is absolutely crucial to understand this the underlying reason Why this change is going to happen because now that we can work down to that and find some a valid basis it's not just AI technology hype it's real psychology it's proven to have been done over and over again our brains want an easier solution and the reason I wanted to include this in the video is because at least for me personally as someone
who likes to think he's logical and evidence-based decisions understanding that there is a factual underpinning to This entire Trend especially in these kind of markets these AI hyper Cycles crypto whatever it is where it's so much hype digging down to this underlying sort of psychological principle and finding real evidence that this thing is going to to be real and it has some sort of backing is at least important for me and I thought it would be valuable for you people who are looking to become chatbot developers to get into this particular space to have this
kind of Thing to go back to and say yes I know that I'm I'm seeing through the hype there is something real to this and with that you can have the convection to really dive straight into something with a bit of evidence behind you and with some real conviction I thought I'd throw this in here as a pretty accurate example of the kind of future we're moving towards just to illustrate this but Iron Man's Jarvis is a great example of the future that we sort of have Laying ahead of us if this technology moves in
the right way so with all of that theory out of the way and we've identified that there's a big thing happening and that businesses have a lot to gain by implementing chat Bots be it through personalization leading to increased Revenue be it through reducing staff costs by automating certain tasks Etc and their customers are very soon going to be demanding these kind of options for interacting with businesses How can we latch onto the back of this trend and build businesses build skills make money and and build a life we love living off the back of
this thing and so the real opportunity for us is that business owners are going to need it but they're going to have absolutely no clue how to do it there are a number of different roadblocks in the way between business owners and being able to implement this themselves some of them are on screen here so chatbots Themselves can be complex you build and maintain or at least initially appear to be complex to build and maintain what I'm going to be teaching you later on in this video is going to make things crystal clear for you
and making complex chat Bots is going to feel like a walk in the park secondly poorly made or bad chat Bots can actually do more harm to a business than good so there's a requirement in building these chat Bots that prevents amateurs from getting in There and thriving around their businesses so there's a level of skill required to create a chatbot that's actually going to benefit the business so again these are sort of factors that are blocking the way for a business owner to be able to do it themselves thirdly depending on your industry business
owners can have all sorts of issues with data privacy and concerns so being able to be the expert and helping them understand this issue and how it Applies to their chatbots and their business can be another way for you to to profit off this knowledge Gap and finally the actual integration of these chat Bots with their existing system so not just building and maintaining the chatbot itself but how it fits into all of the systems of the business can also be complex and and that's something that I'm covering a little bit in this video but
also later in another video that I'm doing so you're going to have all the Information you need to cover all of these points off but these are the questions that business owners have and that's what's stopping them from being able to do it yourself you are going to be the ones that bridge that for them and you have the specialized knowledge to get this thing done for them and allow them to adopt this new technology into their business so that they can make the most of the gains that they provide and I find in these
kind of Situations it's helpful to look back at history to see where we might be going and we have three examples of this kind of thing happening over and over again firstly websites same kind of thing new technology complex business owners don't understand it you could help them to solve it and give them specific knowledge to get them to where they want to go which is their website being able to take advantage of this new technology same thing happens with mobile apps Business owners have no clue how to make a mobile app but you can
help them bridge that Gap but there's benefits for them to receive on the end of it same thing with social media marketing and social media itself businesses knew that there were benefits to be had from it people came in and built a specific skill set or a specific bit of knowledge and then they were able to help businesses and get money in return for that in order to take advantage of the Technology you are the knowledge Bridge bridging this gap between business owners and the kind of gains and growth they want to get out of
these new technologies same thing applies for AI and chatbots and I know what you're thinking I'm not a developer I wouldn't know the first thing about helping them build a website or a mobile app or things like this I couldn't have participated in those the best thing about this current Revolution and this New technology that we have is that it's perfectly timed with the sort of massive uptick in no code platforms and no code Builders so on screen we have a bunch of no code options chat base bot Pro stack make voice flows API all
of these things these are sort of the key elements of building chat Bots at least from my experience these are no code platforms that you can use to build chat Bots and this is sort of the perfect crossing over of this new technology of AI and Bot Builders with this no code Revolution so we have an opportunity as regular people without development backgrounds to take advantage of this and help to be these facilitators Bridging the Gap for business owners there is no excuse why anyone watching this couldn't learn them and couldn't start to be that
person to connect businesses with the sort of Promised Land of having these chatbots built into their business they do all these things For them so with all of that what are your options well from my experience working with students and working with my community members and in my own Journey as well you really only have two routes firstly you can become a bot developer or a conversation designer in order to do this all you need to do is put aside a couple weeks or a month to really Master This Modern chatbot building and and what's
called conversation design nowadays and then Start to sell your services as a freelancer or even better yet get hired by a company who needs these skills as you can see on screen already this is on indeed.com one of the largest job marketplaces in the world and we're starting to see conversational AI designer AI AI chatbot developer Ai and Bot developer these things are starting to pop up with serious money behind them and we're really just starting to see the the beginning of this trend I Believe so if you work on it now and you get
these skills brushed up you can be working as a freelancer doing upward jobs or even just running your own freelancing business a good example of having a no code skill set becoming valuable is the example of webflow I'd say no code website builder but people become webflow experts according to Glassdoor as of February 2023 the national average salary for a webflow developer was 85 000 per year so this Shows you that these no code skill sets can eventually become valuable and you can start to make a lot of money as a freelancer or as an
employee and this was no code for websites now imagine this but combined with the hottest Market in the world right now and something that is really driving a ton of new and exciting value for businesses I'd be very surprised to see these kind of prices if not more as the sort of basic rate for these developers and These bot developers moving forward now option number two is to create your own business that sells AI chatbots and other sort of AR related services to other businesses and this is what I've called the AI automation agency or
AAA as I mentioned in the intro as an AI automation agency you can create powerful AI chat Bots and sell them to other businesses but you can also stack additional value on top of your chat Bots by adding additional automations That integrate these chat Bots into the businesses so it's no longer just a little isolated chat but that does one function it is a a key part of a business imagine a lead generation chat bot that is now bringing you business in and you're sending that information off to other parts of the business you're analyzing
it you're doing follow-ups and things like this so the goal is to package up a bunch of different automation so maybe it's a chatbot with A couple different automation deliverables on the back of it and then you're going to package it up and sell it to a niche you can create a package for a specific industry maybe a chat bot with a couple things built off the back of it you can say that I have an AI automation package for real estate agents and then you focus on that Niche and you're able to sell this
package to as many real estate agents as possible so this is more of the business side of Things I think for Lancing is both great a great way to get your feet wet and to build up a bit of experience and then once you've seen enough of the different industries that are working and the really exciting ones you want to work with then you can start your own AI automation agency focus on that Niche and create like a really awesome package that provides real value for the business and then start to sort of mass distribute
that package to people within That Niche the AI automation agency model is something I created in order to provide the best on-ramp for entrepreneurs into AI because I always sort of struggle to find a way that I could recommend to a friend and say hey this is how you can start an AI this is how you can start to make money the state can start to make a business so I've built it sort of around initially at least around no Code and low code tools so that it's much more accessible And this is the kind
of stuff I'm going to be teaching you later in the video the exact chatbots that we're going to be building are the ones that I sell in my AI automation agency over at Morningside and you're going to be able to take them and start to sell them to clients Etc so I'm giving you all the information you need to to start doing the same things that I am and on screen you can see the kind of transition that I really want the AI automation agency Model too so tight which is transitioning from an AI spectator
and someone who's just looking at watching videos and trying to keep up with the news to someone who's an actual active participant and you're running a business you're speaking to clients you're actively doing things and this is how I've learned so much in the past sort of year year and a half is by actually doing it every day and spectating you're only going to get so Far so that's why I've created the model to be as beginner friendly as possible at least initially to use these low Code and low code tools so that you can
become an active participant and really accelerate your learning and see above the crowd which is really in a position that I am as well and it's really great because you see the opportunities a mile away and you can make much more calculated decisions and you're ready for those opportunities when they come And as I said earlier selling these low code and no code chat Bots to my clients makes up a significant portion of the 120 000 in Revenue we're doing monthly at the moment I don't want to jump the gun and spend too much time
on this now so I'm going to be giving you a lot more information on this model and on how you can get into it at the end of the video where I cover how you can sell chat lots to businesses but if you absolutely we cannot wait and you really want to know More about this right now I do have a two hour long beginner's guide on how you can start an AI automation agency step by step with what nearly 200 000 views on that and the reception has been fantastic so if you want to
jump over it's going to be up here but be sure to come back and watch this because this is sort of the foundation on how you can actually deliver the services and this is part of the reason they're making this is because so many people in my Community were like man I just need a really good tutorial on how to build chat bots in the first place and finally we're in section number two we're going to be covering the different types of chat Bots now this is important to understand because sorting out all of the
different types is probably thousands of different ways you can build them at this point being able to sort them out and at least cluster them into sort of rough groups is important For you moving forward to know what to apply to what situation and what client might need what and what particular type of chatbot is good for which situation now the different types of chatbots that I'm going to cover within this video which covers basically as much as you need to know we have old school chat Bots we have pure AI chat Bots and then
we have what I call Modern chatbots now the first type is old school chat what these things have been around forever They're largely based on decision trees as you can see on the side here sort of like if then if this happens into this if the user says this or if the user presses this button Etc these typically require strict matches and you may have even encountered these on some websites over the past couple of years where you would ask something and it would just have no clue what you're talking about and it's like how
does this thing not understand it's because it's probably Based on strict matching mean you have to say a specific set of words in order to get a specific outcome so in a lot of cases these are going to prefer to have binary choices giving you would you like to do this or that rather than saying what would you like help with and letting you answer in free form and sort of open-ended conversation and I call them old school because they have literally been around for forever really all the way back to 1966 was the first
Chatbot called Eliza it's like a psychotherapist a project someone did at University well this thing had was hundreds or probably even thousands of responses and then when the user sent a question and it had to sort of match up the question with one of the existing responses as best it could so we're here to put thousands and thousands of different responses in this thing and it was all F basically if he didn't put it into the chatbot originally it's not Going to be able to to create it so having to match up inputs and outputs
is a sort of black and white day and night contrast to what I call Pure AI chat Bots and these are basically chat GPT how you're familiar with using chat CPT is the most basic example of a pure AI chatbot you can imagine if your AI chatbots are completely freeform conversation you could chat to it all day you can ask it any question anything you want in the world and it's going to Come back intelligently from that these things run on your endless loop as I said they're just going to go on and on and
on as a complete opposite of old school where you have to have it sort of baked into it and match the inputs and outputs up with these pure AI chat Bots they've kind of done all the baking for you these large language models have so much data it's had hundreds of millions of different sentences and phrases already baked into that so that you Don't have to do it yourself when you ask it a question it's able to just intelligently guess the next token until you've got an answer so you guys are probably very very familiar
with chbt I don't think there's much explanation needed for this but in the case of when we're using it in our own applications and chatbots we build or on other apps we're going to be running on a large language model API so in this case a lot of the chatbot Builders are using the Chat GPT API or the gpt4 API essentially this allows us to use the functionality of chat GPT or gpt4 within our own application so it's basically like calling them up and saying hey I'm going to send you my message can you send
it back can you do that programmatically rather than within the chat GPT interface itself now for our purposes I categorize pure AI chat Bots into two distinct types we have firstly prompted assistance and then we also have custom Knowledge chatbots now prompted assistants are very basic they're very easy to do you guys have probably even done them within chat GPT yourself essentially all you're doing is combining a large language model and a prompt so here you can see I've got instructions.txt whenever you see this just know that I'm talking about prompt so it's instructions to
the model and what I want it to do in this case we're combining a large language model with a Prompt telling it what to do how to behave you're setting it into a mode or a character and it's going to use only the knowledge that's already been baked into the large language model so the easiest way for you to understand what these things are is think back to when you're using chat GPT yourself I'm sure there's been many times where you've given it instructions you are a desk listeners I want you to do this this
and this only respond with this this and This in the same way that you are prompting chat gbt to get to do a specific thing personally a prompted assistant is just sort of the the external version of that we are doing the exact same thing but with an API to do it instead so we provide instructions within our chatbot application saying one example maybe you are a python tutor you are going to teach me how to Learn Python you are going to give me instructions to do this unless you're Going to give me tests those
are instructions that I can give to my chat GPT API and I prompt it into this role and then I can have this interaction with a chat bot that's been prompted into this role now the key thing to know here is that it's using the baked in knowledge of that so you can use any of the knowledge that chat gbt knows and use it off that and have fairly useful and interesting prompted assistant but it's limited and that the information is Only going to be kept at the information that was trained in the model in
the first place and the last Point here is that they can be helpful but they're never really going to be hugely valuable because it's all so freely accessible anything that you do with a prompted assistant can be replicated basically instantly by someone giving it the same instructions so there's no real source that you're adding on top of it you're just giving it a bit of instruction so It can be helpful but in terms of selling it and creating a valuable chat bot for a business especially it's not going to be hugely valuable next we have
custom knowledge chat Bots which I'm sure you're familiar with this is the same as a prompted assistant but you have the added element of a knowledge base so it's a large language model API we're going to cover apis soon so don't worry about that we have a prompt as well for some instructions on what this AI is supposed to do and then we have it connected to a knowledge base of maybe a business's business data so we might have your spreadsheet put in there or you might have some instructions you might have a customer service
document you're creating a custom knowledge chat bot that has access to some interesting data can be very very valuable and there's thousands of use cases really for it not just for the businesses within your personal life there's so Many different things you can do with the customized chatbot knowing how these work and how to build them is such a valuable skill so I'm going to be covering it a time we're going to be doing five or six different examples of customized chat Bots later in the video so stick around for that because they are very
very easy to make when you know how to do them the way I do them but there are so many different applications and once you have that knowledge you can Really provide value to a business in so many different ways now the final Point I've got here is that they are not as simple as you think I see so many wrong things being said about the myths and misconceptions that I'm going to clear up in this video give you the full under the hood look and understanding of the key Concepts and how these things actually
work and that's going to put you head and shoulders above everyone else in this market if you actually know How these things work because when things go wrong you're going to be able to debug them and figure it out so all of that later in the video I I really can't wait to say about that that's my my favorite stuff really some examples of custom knowledge chat Bots are basically customer support chat Bots you have basic AI Persona chatbots that I talk a lot about on this channel I get basic lead generation chat Bots and
basic staff training chatbots now I say Basic because modern chatbots which I'm covering next are sort of the advanced versions which have a lot more features we're going to be building a bunch of basic chat Bots later in the video but then also building some Advanced ones so I thought I'd make it clear the difference between the basic and the advanced ones and the sort of category that they fit into now the final kind of chatbot and the one we're really going to be focusing on learning how to build Later in the video is of
course modern chat Bots this is a term or a definition that I've given to a specific type of chat Bots that use a lot of the modern stuff that we are seeing in the space right now and it typically combines what we've gone over as the old school chat Bots with elements of AIS with the same kind of structure and the way you can lay out a old school style chat but also by adding in a ton of AI features and using AI technology to enhance it it Becomes this really flexible and and Powerful type
of chat bot and as I just said these chat Bots are able to do specific functions or have different features built into them this can perform external actions like pushing and pulling data to a database or calling an API and retrieving some information or sending some away so you can do a lot more functionality and they get out of this little box that these uh pure AI chat Bots are typically stuck in And because of the points I've just mentioned these chat Bots can be extremely valuable to a business and I've got some examples here
on screen of advanced chatbots that you can build you may notice that these are exactly the same as the custom knowledge base chat Bots that we've just gone over the point is that those were basic these are advanced in that you can add a ton more features into it to do a lot more to connect to parts of the business so Anything that you can do with a custom mold chat bot you can basically upgrade and make a ton more valuable and Powerful by using the modern chatbot creation methods now that you've got a clear
idea of the different kinds of chat Bots that you can work with and you can create that's important to now jump into the kind of softwares that are going to allow you to build these things as easily as possible and and sort of give you as much depth as you need to Build the features you want so I've worked with a lot of softwares I do this on the daily I am building chatbots I'm helping my team build chat plus I'm working with students building chatbots so I've really tested out everything and so for my
trial and era I've compiled this list of softwares that you can at least start with feel free to explore your own but the ones I'm going to be covering in this I've used myself and I actually enjoy using it and I was happy Enough to share them with you guys in this video as starting points at least so when it comes to chatbot building I've divided these softwares into three distinct Parts firstly we have prototyping secondly we have what I'm calling complete Builders so these are bot building platforms and then thirdly we have tools which
are extra bits and Bobs that can help enhance the functionality of your existing chatbot to start off with we have prototyping Softwares now there's a ton of different ones you can use for this these three on screen are the ones I typically like to play around with I think there's little differences between each of them that allow you to use it for different use cases and I feel like at least from my experience having these three in your pocket at least being familiar with them is very handy for working with clients spitting up prototypes Etc
so we have chat based Dante Ai and Cody AI we're Going to be covering these later in the video I'll show you how to whip up a quick prototype chat bot and why you might want to do that as well well typing I mean a kind of chatbot that allows you to quickly spin something up it's good for quick demos as I say here on screen you can shoot a little loom video if you're using it to send to clients potential clients and get them warmed up in the idea of a chat bot you can
very very easily take some custom Data put it into these platforms and spin up a prototype this can be very very good for proof of Concepts similar to demos but proving that say you're working with a client this particular set of data I can actually make work within one of these prototyping platforms so you can quickly grab it spin it up and prove to your client proof to yourself and the client as well that this kind of outcome is possible even using these basic prototyping Platforms so if you can do it with one of these
it's very very likely that you can do it and better when you take it over to a complete bot Builder and the reason these are prototyping softwares are not fully fledged at chatbot Builders is that there's typically limited customization on them you can change the styling you can play around with the prompting and things like this but there's a sort of hard and upper limit on how much customization you can Do and that really restricts a lot of the chat Bots you can build here too through that pure AI kind of type right so you
can't get it to these modern chat Bots with a ton of cool functionality ability to call different things so very much contained within a box but still very very useful in in many use cases as well and the key Point here and why I've even mentioned them at all is that these can be very very valuable to quickly spun up a demo show it to a client get Some interest at least prove it to them that it can work as you'll find I've found and my students find as well there's so much skepticism about this
technology sometimes you need to have a very very low and very low risk offer for a client in order to get them to see the value in it so rather than expecting them to pay you five six seven whatever thousand dollars for a complete modern chatbot build that's plugged into all of their business you can spin them up a Prototype for one of these things give it away for free or a very low amount of money and then once I see the value you can upsell them on the full package that's going to do a
ton more and provide a ton more value so each of these softwares are very similar in how they work for chat based you can upload a website URL and take a selected number of pages off a website URL and create sharp one out of that you can also upload documents and it's going to Create a custom modes chat bot that you can talk to and deploy Etc very similar with Dante AI only it also has a YouTube video data loader so you can put a whole bunch of my YouTube videos in and start to have
a customer check what you can chat to there and similarly Kodi AI I put this one in here because it has a cool feature where you can select folders and as you see here we have a website and documents as data to load but then you can also build in folders And then select what folder your custom launch chatbot is going to use so it might sound a little bit complicated but I'll show it later in the video of how we can use this and there's some specific use cases that this can be very very
useful for at least when prototyping then we have what I call complete bot Builders we have bot press and voice flow if you've watched my channel much before you know I've used bot press extensively at least I've been Using it but recently as well as over these two options here they have fairly similar features as you can see here they've got a very similar sort of canvas layout where you can plot out the nodes you can connect them up and you can run all your custom functionality and this is really key because it allows you
to build anything you want so anything you can imagine you can basically find find a way to build it on these platforms giving you huge amount Of flexibility and sort of creative freedom and control so they can build anything from Ultra basic chat Bots all the way to some of the most complex ones you could possibly find in the market right now one of the key differences as I mentioned before between this and the prototyping ones is that you have ability to perform actions and really make your chatbot not just sitting in a little bubble
but allowed to do things outside like connect to the business Database or to call apis and things like this so you can read and write the databases you can make API calls and then another hugely important part is the easy deployment to Major channels so these winning platforms allow you to send the chatbot onto a web chat so put it in a little bubble down the corner you can also deploy it to Whatsapp you can also deploy it to SMS and things like this so plenty of easy deployment and deployment can actually be very Difficult
if you were doing this manually but using these bot Builders and these tools and these platforms allow you to quickly sort of couple different clicks and then you're able to get it sent out onto a cool new way for customers and people to interact with your Bots then finally we have the Third category which I'm calling tools and this includes things like automation platforms such as zapier and make but also things like stack AI which I've Used heavily on the channel before to build really custom functionality so if I build some cool custom AI functionality
that I can't necessarily get done within with a voice flow or the bot press I can create it in something like stack AI I do a really really cool AI application and then I can connect that application and the functionality that I've created to my bot and it can just when I get to a particular part in the bot build or they're in the Conversation they hit a trigger and they can call that app I can pass some data to stack AI manipulate it do whatever I want with it then pull all this really valuable
response back and really expand outside the functionality provided by bot press and voice flow on the automation slab with zapier and make we're going to be covering later in the video as well basically these are very very powerful platforms that's going to allow you to integrate the chat bot with The rest of the business and basically do a whole bunch of cool functions again extending the functionality of the chatbot Beyond just this little bubble that it works in but being able to pull data from it and do things with it and Trigger events within the
business like capture leads and do all sorts of stuff trigger email and SMS flows all that really really fun stuff which I love doing I'm going to be covering later in the video as well so we have a lot to Get to when it comes to key Concepts today we're going to be talking about knowledge bases first and foremost which is hugely important to understand and the real complexities and what's going on behind the scenes with them then we have a brief touch on fine tuning then we get into prompting which is a hugely important
aspect of chatbot building and tent classification chat history and deployment we're going to go through each of these right now and be sure to Have your notebook ready because there's a lot of information to take in without further Ado let's get stuck into knowledge bases just to be clear when I say knowledge base I'm referring to the element of the custom knowledge chatbot that we were mentioning earlier in the video it is a key element in creating a custom mod chat bot and knowledge basis where the knowledge is stored really so a knowledge base is
a database of text and numeric data chunks stored by Similarity a chatbot can retrieve a small amount of information most similar to the a query from that database then only this info that was retrieved from the database is then passed to a model such as the chat CPT API to help it generate an answer to their question now not all of the info from the database is going to be used just the chunks that were retrieved that's about as simple as I could get it to make it fit on the screen and we're now going
to spend a Couple slides just really breaking that down the appearance first reality when it comes to knowledge basis is really important to understand initially and that with a correctly set up knowledge base is and you've done all the work and you've managed your parameters correctly you can create a chat bot that appears to know all of the information that you've given it to know but in reality this is not true it does not know all of the information and here's why the issue That we run into when you're trying to make a chatbot know
all of the information or an AI in general a large language model is that we've run into the similarly age-old issue now of language models is that you hit the token limit we have this token limit pane as you may know chat gbt when you're using it personally has token limits is you've given too much the message is too long you need to reduce it and try again it's the same thing When you're using these models programmatically when you're calling the apis there is a fixed limit on the information that we can provide them with
in terms of how many tokens so here on site you can see GPT 3.5 turbo which runs chat CPT there's a limit of 4 000 tokens and this includes not only the message you send but the information you get back so if I send 3 000 tokens and it tries to send back 1500 tokens it doesn't work it's over the limit and Because of these token limits it is impossible I mean it's possible if you have a very small knowledge base but if you have a large knowledge base with like an entire encyclopedia inside it
it is impossible because of the token limit for the AI test look at all the data and really see it all because you can't put the entire book and it can only understand what's in the four thousand or eight thousand tokens that we send away to it so it's impossible for it to Look at everything because it's more than 8 000 tokens and so in order to get around this token limit the issue we need to set up a retrieval system to get the most relevant info to answer our user query and only that we
only have 4 000 tokens if we're going to answer users question we better make sure that the information that we are putting in those 4000 tokens is the most relevant to what the users asked this is the key to understanding this whole thing and so How do we set up a system that can retrieve the right information to answer the specific user question well the first step is what's called chunking and that is basically say again using the encyclopedia example I put an encyclopedia into my knowledge base what I'm going to do what the system
is then going to do sorry is chunk it into into sets of text using the encyclopedia example again say I take my encyclopedia and I put it through our system what It's going to do is take that entire book or the entire sort of thousand thousands of pages that you scroll through on your computer and it's just going to chop it into what are called chunks now these chunks can vary in size and you can play around with the different sizes of them and how much they overlap and things like this but to keep it
simple imagine we're chopping them into chunks of text this big maybe so we've got one chunk two chunk maybe Three chunks per page or something like that and then once we've chunked encyclopedia into thousands of chunks of this size we're going to take all of those chunks and put them into what's called a vector database I don't want to get too technical but on screen is an image that kind of allows you to understand how the effective databases work it's not a normal database and how you have like a row like a spreadsheet and things
like this it's more of like a Box and you can store the information all of your chunks go into the vector database and they're stored by this similarity of the meaning so there's a using the bidding script that's going to essentially identify what the the chunk is about so every chunk we're going to put it through this what's called an embedding model don't worry about that it's a bit too technical but we put it through the embedding model and then it's going to basically tell us what This is about and then when you put it
into the vector space it's going to sort everything by what's semantically similar to each other so on screen you can see here we've got things like barrels and BPD barrels of BPD have been put close together and the closer they are the more similar the meaning is and as you can see here we also have oil producers in China and things like this it's all about encyclopedia are going to be put into our database and all sort of Assembled when they're closer to each other if they're more similar so one on an elephant might be
right next to a rhinoceros and so on so and then something about cars and motorbikes might be in a completely different area but nearby as well it's important to understand this that they are stored by their similarity they're clustered and they're close to chunks that are most similar to what their content contains firstly you'll recognize our Vector Database is up here this is filled with all of our information so it starts over here on the right with a user query so say I'm using our encyclopedia custom knowledge base and I say how much does an
elephant weigh it's going to take my user query it's going to run it to the vector database and it's going to basically have some instructions imagine as someone who's operating the Victor database and they read how much does an elephant weigh and it has coordinates of Where the elephant would be and he goes reaches into this box and grabs okay elephant is over here I'm going to grab anything around there so it will pull back maybe five chunks from our encyclopedia that are closest to where elephant should be in in the Vic database so we
pull that back out now we have five chunks that are highly similar to what we've asked so it might be an elephant or an ostrich a giraffe anything that's kind of similar to what We've asked about it's going to pull those information chunks back and then we have what we're called the most relevant chunks here on the side then we have the final crucial step of actually generating an answer based off this information so here we can see we take the most relevant chunks and we pass them over to what I'm saying is an AI
plus a prompt and we're also going to be passing this AI plus a prompt the user query that we said originally so how Much does an elephant weigh is our original query we're going to ask a bigger database for anything related to that we've got the most relevant chunks back regarding elephants and how much they weigh and we're passing that to the AI but we're also passing us the original question so the AI as you'll see in the next slide is going to know one what we're asked after and it's going to have these five
chunks of information that should be able to help But answer the question so this in particular deserves a little bit more investigation so we can take this to the next slide now there's some code on the screen here don't freak out you don't need to understand stand what's happening here but basically when you're actually making calls and you're using these apis and you're talking to the chat GPT API programmatically you have what's called a system message you may be familiar with this already from the Playground but essentially this top line is called the system message
and that's the instructions you give to the entire chat experience so all the messages that come after are going to be constructed based on what the system message is so in the playground you may be familiar with it on the side panel you have your system message so in this case it says you are a helpful assistant we could say you are crusty the clown speaker across you the clown and make jokes and then Whenever we send a message here on the user side of things which lines up with these also by the way it's
going to take into account the system message and then it's going to answer and give us a response so that's what I mean by a prompted AI what we have is basically we need to give instructions to the AI on how to use the information we've just got back so you almost got the information about the the elephant but we need to now generate an answer using The AI So the instructions that we give in the prompt to this AI is going to directly determine the until we get back so if I tell this thing
and our system prompt to deal with this elephant question you are crusty the clown make jokes and answer questions that the user asks it will eventually give us the answer to our question but the instructions told it to speaker across to the cloud so that's why this prompt and really the instructions you're Giving to the AI on how to deal with this information is so key and it's one of the major and most important parts in the whole custom knowledge based chatbot system or at least using a knowledge base so also here it's important to
note is how these actually come together so yes I said we are passing the user query we are also getting the chunks from the vector database but when it comes to actually sending it off we are inserting the chunks that we receive say this Chunks bit is replaced with five of those third of a page chunks we've got and also the user queries so we can tell the AI that you were going to be given chunks of information that may be relevant to what the user is asking and we're going to provide it in the
prompt so then when we pass the chunks in let me also pass to use Aquarian it's like okay I need to answer the user query with this information up and given and you can tell it in the prompt to be Strictly using the information that came back there's a lot of different things you can do here but it's really important to understand that this AI plus prompt is really the magic step of combining what we've retrieved from the vector database aka the the most relevant information to what the user is asking with combining it with
the user query and then asking the AI to generate us an answer going back to our full system here once it's generated the Answer it sends it back to us as the response that we see appear in the chatbot so I hope I've been able to communicate that as clearly as possible it's really important to understand that the answer is actually being generated based off the information so it's only going to have the information that we got from the knowledge base you can't that's what I'm saying we only have this 4 000 token limit we
don't have the entire 10 000 or 100 000 tokens in the Encyclopedia we can only work with the information we have and this is a common misconception about knowledge basis is that oh you just throw it in there and it's going to know the whole thing know what it's doing is intelligently retrieving the information so that it's able to look like it knows the information really all it knows is the most relevant info from your knowledge base for regarding this specific question so I thought I'd throw in an Analogy just to help you understand this
a little bit more so you are going on to Google and you're doing a Google search so yeah you search up how much does an elephant weigh so the Google search that you may do how much does an elephant weigh goes to Google search algorithm and it Returns the most relevant results first so you get the first page results of Google and somewhere in there is going to be the answer to your question right and then what they do is instead Of passing it to an AI to dig through and check we just pass it
to you the human to dig through and you know what the question you were originally thinking of is now you just have to pick through all of the information and find a bit that gives you the answer in the exact same way we have the system set up where the AI is given the most relevant search results from the vector database and then it's given the original query okay what were we actually looking for In the first place and then it figures it out looks all the different chunks it's given and then gives you the
response that's the best analogy I could come up with I hope that's made something click in your head now for those of you who want to go extra deep I'll just put this on screen for a second here let's stay a little bit more in depth as they how things go I don't expect you to understand this but this is a system known as a retriever Generator model so we're retrieving the information and we're generating the answer you can pause on screen and just take a look at this this is missing a couple steps but
if you want to know how this really happens step by step you can pause it take these notes but this is really what's happening under the hood with a custom knowledge base and as I said earlier there are some common myths and misconceptions that I must debunk in this video just for the sake of my own Sanity from hearing it all the time but also for you as someone who's soon going to be speaking to businesses and speaking to potential clients if you're going to start selling these you need to know this because you could
sound like a real idiot if you say it in the wrong way so number one you are not training a model there is no training being done training a model costs millions and millions of dollars and huge ridiculous amounts of computing power people like You and I do not have the means to train models unless it's some small maybe with a big rig that your friend has I don't know it's basically not on the cards for any of us so training a model we leave to the big multi-billion dollar companies to use their resources and
train models with new data and that really bakes in and information into the model you are also not fine-tuning the model A lot of people use this term you are not fine-tuning a model when you Have a knowledge base all you're doing is making the right information accessible to the model to use and answering its question and sort of an intelligent and flexible system I think the terminology of fine tuning can be a little bit of a gotcha because in many ways when you're working with these kind of chatbots you can go to the client
and say okay I'm going to fine-tune the responses a bit but that doesn't mean you're actually fine-tuning the model by By its actual definition it could mean just going back to the drawing board and tweaking The Prompt a little bit and how the instructions you've given on how to generate the answer you're going to change those instructions a little bit and fine tune it in that way to to really get the exact kind of outputs that your client wants and finally you are not teaching the model anything it's not learning you literally just putting the
right information in front of it to Answer the question at the current time and then it goes back to being just as dumb as it was before so it's you're not teaching you're not fine-tuning you're not training I thought I'd just get that cleared up right now I will give a pass to my one situation in which you can utter these words and that is when you are talking to a client who is clearly not Technical and he or she may struggle to grasp the real Concepts that I've just taken you through now and Explaining
that to them is probably going to do more harm than good it does sound better and it is easier for them to understand if you say we're going to train a chat bot on your data and and that's that's the only time you can probably use it but I will just be very careful that you don't say it to the wrong person because they'll look at you is pretty stupid if you say that before we wrap up the section I thought I'd throw one more analogy your way and That's to help you understand these knowledge
bases and how the difference between chat Bots learning and you just giving them the right information at the right time is imagine if you were to ask me a question about anything we've just gone over in this video I would be able to give you an answer in a second because I know the information that's already in my head but if you ask me a question about your business's customer support documents I wouldn't know how to Answer but unless you gave me the document so what I'd ask you for is hey can you give me
some pages or some documents that are most relevant to answering this question you have about your customer support you can give me the five documents and I can go okay that but there that but they're okay yeah I know the answer the answer is this that is the simplest way that I can explain the system you're giving the AI a couple documents to answer the Question that's checking over all of the information and it's going okay this is the answer the next key concept we have to cover is fine tuning are we going to blast
through this one pretty quick but basically long story short chat models aka the open AI API models that we use for basically everything in chatbot building at the moment they cannot be fine-tuned it is coming soon they say they're working on it later in the year in 2023 and the ones that you could Fine-tune the older models which you were able to find tune that weren't these chat models are actually being depreciated so you can't fine tune the old ones and you can't fine-tune the new ones so fine tuning is basically off the cards if
you want a little bit of technical on how these fine tunes actually work it's narrowing down a model to a very very specific use case so for chatbots this is basically useless you're teaching it to identify a Pattern and then respond to that in a specific way for example you can fine tune a language model to identify a bunch of information scraped off a credit card statement and teach it to only identify alcohol expenses if you gave it enough examples of the information and then the output you wanted you could fine tune it to be
able to instantly pluck out alcohol expenses or something like that so it's very very small use case and again an important Point to mention when it comes to fine-tuning you are not adding new data to the model what you were doing is teaching it to identify a pattern and respond to that pattern in a specific way it's pattern recognition and pattern responding really basically forget fine tuning it's barely used anymore and it's not really relevant for anything you're going to be doing in sharper body next we have prompting which in its most simple definition is
giving instructions To the AI on what it should do so in our case for chat by building we're typically prompting the AI into acting a certain way or being a certain role or maybe you give it a name you say your name is Dave and you do things for this business and this is how you do things and this is basically the instructions you want to tell the AI on how we want to process our information how we want to respond to users what things to do what things not to do now prompting Itself is
a huge huge topic and I've actually done two complete videos on this which are long enough in their own right so I'm going to link them here if you guys want to check them out it's going to be up here it'll also be in the description I can't cover it within this video because there's seriously so much groundwork you need to I know but trust me these videos that I've got cover it from basically zero all the way to very very Advanced and if you're able to Watch that after this and then apply that to
the chatbot building you do you're going to be in a very very good position so everything I know and use on a daily basis is in here and then you're also able to use the prom templates and and the actual prompts I'm giving away later in this video alongside the chatbots we build so with this knowledge and then having access to the templates that that I provide and we find in my community and you're going to have all The prompting expertise that you need so definitely check these out next but just know for now that
prompting is giving a set of instructions to the AI in the same way you'd give a human employee a list of instructions this is what you to do in your job so that's about it for property next key concept is called intent classification and this is a huge huge thing that you need to understand in terms of modern chatbot building so intent classification is defined as the Task of taking a written or spoken input and classifying it based on what the user wants to achieve now this is best explained with a diagram so here on
the left we have a chat bot let's say for these purposes it's a e-commerce customer support chat bot so as soon as the user sends a message it's going to send it directly to the intent classifier function now this can be typically done with on the bot building platforms like voice flow and Bot press It's going to classify or attempt to classify what the user is asking for and then if there's something that's been set up on how to deal with that specific attempt it's going to Route the query that way and then kick off
a bunch of functions so basically if they mention something specific that falls into a category we can do a special thing so this is great for keeping the chat open and basically allowing people to send whatever they want instead of giving Them options so you can either give them buttons and say I would like to do this I would like to do this or you can just say hey how can I help you and we're going to figure out what you want using our AI system so for the e-commerce store example we could have action
one here action two and action three I've actually got it here and 10 classification example say we have a message coming from the user so first the chatbot asks hey how can I help you Today leaves the floor open and then when the user sends a message we're going to run it through the intent classifier we have a couple different intents set up so firstly they've mentioned something about shipping like Hey where's my package or I don't can you update me on the shipping in that case if it fits that bill it fits the intent
of wanting to know more about their order or whether you're tracking or what the status of tracking is then We're going to send it to this one here and then it's going to kick off a bunch of functions where we actually check that information for them so in this case it will go up and ask for the order number capture the order number check the system and then give them the answer furthermore if if the question that they sent was like Hey I've really enjoyed my products it'll arrived just on time I just wanted to
say thank you I mean no one's really going to say that but if Someone did send that to a chat what sort of spontaneous obviously we can have an intense setup that is I've enjoyed my product I'm really happy with the product shipping was fast anything related to that kind of message we can say okay we want to identify that and do a specific thing and in this case the chatbot could respond by saying hey I'm glad you liked your product and you go like that shopping experience would you be interested in leaving a review
so This is two examples of being able to catch something that the user has said while still leaving the floor open but we can sort of pluck out specific things that we're looking for and then if that thing is in the message then do this and it allows these AI chat Bots this is how you go from a pure AI chatbot to more of a modern one where you're able to open the floor and let them talk and ask general questions but we can pluck things out and do things uh that that Sort of pure
AI chatbots card and the key thing here is that you can set up a default where you may just have a knowledge base on General customer support FAQs and if none of your intents are triggered say there's nothing related to shipping there's nothing related to how much you love the product so it's not going to be route to these specific functions we've set up we might have two we might have 10 we might have 20. if none of them are triggered then We can just send it to a regular knowledge base with our customer support
where we say we've prompted our AI into thinking that it's a customer support agent and it's just supposed to respond to bigquery so if the customer asks for anything outside of what we have built into our intent classifier we can just send that to a custom odd chatbot and answer the question for them normally I want you right now to just to really dial in the understanding of this Concept is think of some kind of chat bot maybe it's a customer support chatbot like this or it's a staff training chat bot or it's uh whatever
kind of chat but you can imagine in order to really drive this concept home I want you to take your PIN right now on your notebook and sketch out the sort of diagram here and pick an example of a chat bot for example a staff training chat bot for a construction company which has all of their staff trading Documents included in it what we want to imagine is if we have a couple different features so say a staff training chart but what features might they want in the okay they might want something where the staff
could request time off or they could report an incident on site or they could this and then look at that so take your idea and plan out what kind of features would actually be cool uh within a chatbot for us to be able to do actions for the business and and sort of Set it up and connect it to some external things and actually do something else on just the chatbot Now intent classification makes up a major part of what we do with the advanced chat Bots later in this video I can't wait to show
you guys those really really interesting similar examples to this but you get to sort of see it in action as well so just a recap intent classification is a very very powerful tool for Chapel developers currently it Uses AI to analyze what the users ask for and from their trigger different actions to allow it to do external functions or do additional features and this of course as I just mentioned is a key component in modern chatbot building next we have chat history which within chatbot building can be a quite a difficult subject or quite a
difficult thing to get to work properly within your application so I wanted to touch on it here and just give you a little bit Of a background on how it actually works when you're doing a different chatbot applications basically what the system for handling and remembering context within the chat is so as we went over and the knowledge base section we only have a limited amount of tokens to work with and at any given time your the model that you're using for your chatbot underneath can only look at 4 000 tokens or 8 000 tokens
or 16 000 depending on which one you're using so in order for It to know about the chat history context we need to be giving the chat history to the AI including it in the prompt that we send away on screen we have a little bit of an edited example of what you just saw earlier which is the same prompt that AI concept where we're giving the AI a prompt we're telling it what to do but then we're also retrieving the most relevant chunks but in this case we're also adding in a section for the
recent messages this is A way we can include some recent chat history as context in order to help it make its output so in this case you could prompt the AI and the system prompt up here and include it this is going to be a lot longer than it is here I obviously don't have room to fit it all in but within that system prompt you could say you are also going to be provided with some reset chat history use the messages provided to be aware of the recent chat history and factor that Into your
answers so basically instructing it you're going to be given chunks which is going to give you the context for how to actually answer the question the information that you need then you're also going to be given a a section of chat history which is going to give you a bit of context in order to properly answer the user's question given the the past messages as well now given all of this information answer the user's query at the bottom here so a Little bit of a clear way of showing this is to put it right here
so imagine each time that we actually send a message off to the AI say we see the message on our chat bot every time we ask something this is essentially the the stack that's being put together before you send it away to to the llm to create the answer so we have firstly the system prompt which is going to be our instructions at you are Dave from Jerry's gym or whatever it is as we said That's going to include the sort of UI to use the information provided in the context to create an answer be
aware of the chat history provided you should not do this you should do this Etc that's a little bit at the top then we can add in the most relevant chunks from our Vector database for my knowledge base being put in there as context to help provide information to answer it with then we can add in a section of recent chat history and then finally we can add in The user query and then we bundle that all up and then we send that away to the in this case probably the chat GPT API in order
to get that responder to and we send that back to the user as the final result so this is kind of chat history knowledge bases prompting everything built into one here on the screen so I hope this makes sense because this is kind of the The Ice Cream okay on how things are really being assembled behind the scenes every time a user sends a Message next key concept you need to understand is deployment now this is a quick one I just wanted to make sure everyone's on the same page because I will be using the
term quite a lot we have deployments is how you're going to get your chatbot out how is it going to be exposed to customers how are people actually going to use it now it's all good and fine and dandy to build these chat Bots Within These bot Builders as we're going to do later but when it Comes to actually providing value to a customer it's very very important where is this thing even going to be is it accessible enough are people going to be able to use it so we have a couple of different deployments
on screen web chat of course is a very very popular one you can create a chatbot on these platforms and then put it as a little bubble in the bottom right corner and people can access it you can put it on specific pages of the website or you can put it On every page of the website or you can even take these shop lights all of the time and put it in a sort of full page chatbot example and still put it on the website then we have WhatsApp which is very very useful very very
simple deployment and works very well for some Industries and not so well for others some Industries are very very prone to using WhatsApp and if you can deploy something there it's very very basic it seems basic but you can build a lot of Cool functionality into it then we also have Facebook Messenger so you can deploy to Facebook Messenger and you can also run lead you can run Facebook ads to this messenger so it really ties into a completely different side of the chatbot realm when you can actually run ads these chat Bots and starts
to lead generation and things like this and then finally you can use twilio for SMS and and be able to chat with these chatbots Via SMS as well so this is really just Scratching the surface you can do Discord slack all these different things basically anything you can imagine you can deploy to but these are kind of the main ones you'll be looking at a lot of the time now the key concept section would not be complete without covering a little bit more of the jargon first off we have apis apis is an acronym for
application programming interface a software intermediary that allows two applications to talk to each other so That's obviously the technical explanation for it now in order to understand apis which are a huge part of chatbot building you're going to be calling apis within your chatbots to perform external functions to connect your chatbot to outside data sources and to go to zapier and make and things like this so it is absolutely crucial that you understand how these work because you are going to be building them and this is about as Technical and as Cody As as you're
going to need to to know because you're actually going to be using these pretty regularly within your chat bots so in order to understand apis you have to understand how the web works really which is by requests and responses now when you run up a web page say you search up google.com what you're doing is sending a request to Google server and what they're doing is sending the information for that web page back to you and now that's loading on your Screen as a web page now that doesn't necessarily need to be an entire web
page the URL in your browser is not just for finding websites that's also kind of like a phone so you can make contact to different available resources on the internet so when you type in a specific address you're basically sending a request through that address with hey I want to get some information from you and it can send it back if it's a valid URL and if you've provided the right Information for it so the way I can explain requests and responses in this case is that say you are sitting at home and you want
to order some pizza you need to First find a pizza shop and you need to find the phone number now their phone number is exposed to the world that they can call it's open and available for people to make requests to so you can call the pizza shop up on their specific number and you can tell them exactly what you want and then in No time you're going to have the pizza arriving to your door as the respond so when you replace the values instead of yourself you have either your application which in this case
will be a chat bot or your browser on your phone you make a request to a specific URL or what's called an endpoint and depending on the URL or number you've called there's a set number of things you can get back from them you need to include loot in your in your call and your Request to the specific URL or number what you want back from them and then if everything checks out you know say you have the right authentication you've given the right password to get in which is where API Keys come in the
server is going to process that request and send the information back and that's of course when it comes to the web there's information moving back and forth not pizzas of course so you're making requests to specific URLs or endpoints In this case I might have the liamotne.com API and you may be requesting to get from my photos maybe I have a gallery I have a photo section you may be requesting the first photo or the photo of ID number one you're requesting hey can I get photo ID number one and then it goes to the
server okay yep that's all checks out then it's going to send that data back to your application or to your computer and maybe within that data there is a URL to The actual resource location of my photo when it comes to API calls within chatbot building a good example is uh capturing lead data so say your chatbot is chatting away and a client is tapping away on it and your chatbot sends a message and says hey would you like us to contact you via email or you get in contact with an expert they go sure
and then you can send hey please provide your email and your first second name and then once you have that information What you can do within your Bot Builder is send that information off to sapier and zapier is going to process it or your fuse make.com or something similar it can process that and send that onwards to a Google sheet or to put it into your database whatever you want to do with it but in order to send this data out of your chatbot you need to make a request so what you're going to do
is you're going to be provided as you can see here zapier is now the pizza Shop and it's going to give you a link that you can send your data to so what you're going to do is in what's called a post request is different types of requests don't need to go that deep on it now but a post request basically allows you to bundle up a bunch of data in this case it's going to be maybe the timestamp of when the lead was captured maybe the first and last name maybe the email of the
person it's going to bundle that information up and it's going to Send it in a request to a post request to zapier to your specific endpoint URL are they provided you to so they're basically listening okay we're ready to receive your information when you send it they're going to go okay great we've got it and then they're going to send you back a response and basically approving it in this case and they're not sending any data back necessarily to you that you're looking for but they're just sending back okay yeah everything Was good on our
end we've received information and everything's been processed okay so what you'll get is an okay response and this is what you receive back a lot of the time in the responses from sapi and from there on the zapier side of things when you're setting things up you can basically pluck out the information that you sent in that sort of bundle of information you sent where does apia you can pluck out the name you can pluck out the the Timestamp and the email and you can set that to different forms or different columns within your Google
sheet or wherever you want to send it and so basically you can analyze what information you're getting back every time and tell it what to do with it so it can do its job another easy example of an API call within a chat bot could be part of your chat bot needs weather information and when the user says give me today's weather you make a call to a Weather API you request a specific date range of information say I want the weather for this week and it's going to request it it's going to pull the
information back and then you can use that within your application so it's all about requests and responses sometimes you're sending information away sometimes you're asking for it in return but it's all about the sort of back and forth and the information you're sending and the information they're sending back Find a bit of jargon I want to touch on is Agents these things obviously had a big explosion of growth when our order GPT and things like this came out and I thought I'd just do a little bit of an explainer here just to make sure you're
on the same page and the difference between a strictly a chat bot and an agent so the official definition if there is one an agent is a program that is designed to perceive its environment make decisions and take actions to Achieve a specific goal or a set of goals an agent operates autonomously meaning it is not directly controlled by a human operator now the key word there is autonomously it's best to explain this with an example so say I have set up a travel assistant agent this is the guy in the middle looking happy there
an agent's job is essentially to take a task break it down and analyze it into all the different different parts of it and then start to do those one by one in Order to do a sort of multi-step and more complex task one example of a question that I could ask say I made an agent that's a travel related agent it has access to Expedia all these different sites sky scanner and can it has access to a bunch of tools now this is important we'll touch on this later but the input is going to be
something like hey plan me an ideal vacation in Europe for a week in July now the agent is going to take that in it's going to Analyze it and say okay how do I start to break this down first I need to look up what are the best places to go in July where are the good places to go in Europe and July okay I'm going to first do that and so once it's got this list of to Do's that initial list of first I'm going to search out the best places then I'm going to
narrow that down to three then I'm going to analyze those three against the original question then I'm going to check Expedia for any Hotels or things in the area then I'm going to go and book flights check flights as well and basically once it has this big list of to Do's then it goes on to using the actions that you've set up available to the agent so in this case it might use Expedia API or my just Google search in general ask for where's the best places in Europe once it's got a couple ideas you
might start checking sky scanner for flights okay what's uh what's the cheapest flights or what's The best flights where can I get direct flights if it's going to work so having a bit more information on the input in terms of I want it to cost around this much money I don't want to do any multi-stop flights those kind of bits of information would make it a lot better but once it's gone through this process and it's actually pulled this information gone through its list of to-do's and then it can create a final answer and really
evaluate that so it Will check the final answer since you've gone through so many steps and it'll go back and say have I actually answered the question and then if I can confirm it will give it back to you so agents they do a lot of different tasks they can do step-by-step things and think on their own really and you get need to give them access to a bunch of tools really they call them tools there's different words for them as well but essentially a bunch of different Functions that they can do a bunch of
ways that they can interact with the real world and with the internet at large in order to get something done so when it comes to chatbot building you are not going to be building agents strictly that is quite advanced stuff but what you are going to be able to do between is connect an agent to one of your intent functions say if we had a chat bot and the the user said the same query I plan your vacation in Europe for A week in July you could have an intent classify setup where if they ask
for some kind of plan you could pass it to an agent that you or someone else who created it can run through all these complicated steps send it back via an API call and then you can get that answer to the user so if you can create an agent or you can find ones to use data external and you can include them in the functionality of your app and that is one way that you could Potentially use agents and your apps and with that done I want to give yourself a big pat on the back
you have completed the theory you have completed the explaining the opportunities the uh the different types of chat Bots everything you need to know you now have the foundation that we can build on top of with real skills you've done your college now you're going to go into your apprenticeship you're going to actually start putting it into practice and Getting some real world experience so I'm really excited to jump into the next section which is building these chatbots live maybe go take a bathroom break refill on the coffee because we've got a lot more to
go here so let's get stuck into okay so what are we actually going to be covering within this live chatbot building session so firstly we're going to be doing some prototyping using the prototyping tools as I mentioned I'm going to show you how to quickly use Chat based Ai and also Kodi AI I'm giving you a little bit of background on how you can use each of them and what those sort of workflow is with them then we go then we're going to be jumping into voice flow which is the bot building platform we're going
to be using for all of the advanced builds in today's video and I'm going to be going over a bit of an introduction and orientation so you're familiar with everything and then we'll jump straight Into building the three different types of chat Bots now it's not just going to be three chat Bots it's going to be two versions of each of them so a basic customer support and an advanced customer support a basic staff training and an advanced staff training basically generation and advanced one and so on there's a lot to cover I'm not going
to waste any more time talking let's jump straight into using these tools so the first prototyping tool we're going to Use is called chat base all the links to all the tools and platforms and softwares I'm using in this video going to be available in the description so just head down there you can find links to all of them anyway we're going to jump onto their website you of course are going to have to make an account once you've created your account and signed in you'll be greeted with this page here once you're logged in
you're able to go to the my chatbot section and Then click on new chat bot and creating a custom knowledge chat bot which is what we're doing here it's easy as literally dragging in a file or dropping a link to our website and so on so in this case I'm going to show you how to create one with a file so in this case I've got some different documents that I'm ready to put in that we're going to be using throughout the section this one is a customer support template or handbook really for a gym
called Jerry's Gym which is an imaginary gym that I've made up and got a document for so I've uploaded this it's ready to go it's within my 400 000 character limit depending on your plan you can add more context so when you go to a website if you scrape a really large website you might need to upgrade your plan in order to be able to chat over everything but in this case Jerry's gym stock is uh well within the balance within our free plan so I can create my chat bot let's Get this thing out
of the way and then just like that within a few seconds we have a customized chat bot for the serious gem document so I can say what are your PT plans price and it's nice and snappy and click that's one of the reasons I like chatbase uh it's so quick when you're doing this prototyping it's really good to see and you can say what about normal membership prices that already gives you a pretty good idea of how this thing works the key thing about Chatbase is that you can go to settings you can scroll down
and then you can start to play around with the prompt now prompting as we've gone over when you have a custom launched chat bot The Prompt that you provide is really important because it's basically the instructions on what to do with the information that comes back from the knowledge base I can tell I'm here to speak like a pirate and he's going to speak like a pirate or I can give it Some very clear guidelines on if you don't know the question then do this if you do know the question then answer in this way
or these are your goals this is what you're supposed to be doing so I actually have a prompt all written up and ready to go for this so I'm just going to paste in the prompt that I've written before I'm going to have this prompt available on my resource Hub along with basically anything else I'm going to mention will all be available On my resource type you can sign up and get access to all of it they're all completely free but this prompt is very good iPhone for customer support it gives it a name it
tells them what the objective is it tells if it's in doubt refer to the support email and so on also things like how to format links so if you're going to have links in your knowledge base it's important to be clear about how you want it to reply sometimes you can get it to put it in Markdown and essentially put a a live hyperlink into the chat so that's important to to have a bunch of these things in there so feel free to grab that prompt template and use any of the other ones that are
mentioned in this video it's going to be on the resource Hub in the description then once you've set this up you can really do whatever the platform will out so in this case we can go to Integrations you can add it to a slack workspace but more so for chat Based specifically as I said this is very good for prototyping spinning up a quick demo sending it to a client so what you can do is you can embed on a site so you can go save changes we can just skip the domains necessary and then
we get these two little little scripts here so this is a iframe that can be put onto a website that's going to display the entire chatbot interface but then if you want the chat bubble you can copy this and add it to the bottom so I can Copy this and head over to brackets now this is what you'd give to your client in order to put on their website and test out so I've just got a little demo site here that we can spend the test up on as you can see I've just pasted in
the script tag at the bottom of the website again don't need to know too much about this but this is how a client would be able to use it you can pull it down here and I can start to chat away with it what are your PT class prices And there you go now this is how you can prototype and send it to a client this is why I like chat based it's so quick and easy to get going I've been able to prompt it so it's manipulating my information how I want it to and
then I can even give this to a client and they can put it as an iframe in their website Etc you can also share it which is probably the easier way to send it to a client so if you click the share button and hit made public we can copy this Copy new page and then we have the chatbot as well um where is your Boston location and there we go it's replying and there we go it's all working nicely one other feature you have within chat base is the actual design and the layout of
things so you can come to this chat interface you can play around with the theme light theme dark theme suggested messages user message color and things like this so if you want to match it to your clients coloring in the way that Their website is styled then you can do the same there so that's chat based really nice and quick and dirty great for speeding up prototypes as I said and that's about it for chat based feel free to use it as you wish to play around whatever you need to do next we're going to
be doing some prototyping on Dante AI it's a little bit different to chatbase and if you want to follow along there's a link in the description some of them are affiliate links so if you want to Support myself and the channel and some of this free stuff that I do feel free to use the affiliate link if not I'm sure you can see the URL on screen once you've created a free account you should see something like this just a basic dashboard here what we want to be doing is creating a custom knowledge chatbot to
spin up a prototype with here so what we need to do is create a new knowledge base over here on the left hand panel click create new knowledge base now as I Said in the earlier stages of this video we can either upload video URLs and add sort of 5-10 different URLs and it's going to pull the information from that video from the YouTube video sorry or we can be providing a URL to our website here so I could put my personal brand website and ask questions about the information on the site hopefully for simplicity's
sakes I'm just going to upload the same document Jerry's gym and I'm going to create it and then I'm Going to name this Jerry's gym and then I'm going to create the knowledge base so there is a limit on how much you can do within this a free tier so when you're doing this you may want to keep your documents fairly small otherwise you're going to have to upgrade I got in touch with the team and they've upgraded me to the professional one so I can show you guys some of the other features they have
but the key thing that I've noticed when I was playing around with this and Prototyping on my own is that the resume limit to the size of the documents you can put into your knowledge base on the free tier so if you're testing this out just want to check out how good it is and if you like it or not be sure to use a document that isn't be on that limit where you're going to have issues but once we've created this we have the knowledge base on the left we can choose to speak to
it now we can say what uh your PT class prices And then we have for getting the same sort of responses I do like the the strength of chat history here so if I say what about for normal memberships so here I've tried to ask a sort of contextually aware question and ask it what about normal memberships when I've asked for the price of the PT classes before that said that there's no information about the cost of a normal membership so just picking up on the chat history context however it's Mentioned that memberships are paid
multi in advance Etc now let's try it again can you provide the link and there you go so this is one of the reasons I like Dante AI because the uh the chat history is actually really good an issue that I often have with platforms that I'm trying out is that when you're asking contextually working questions it doesn't pick up on what you're actually talking about and chat GPT is absolutely our enemy in a lot of Ways with this when you're you'll find very quickly when you're dealing with clients that a lot of the time
they will try out your demo or the product you've made for them and they'll come back and say this is like nothing at all like chat gbt it's not doing anything and it's because yes we're using custom watch bases it's a hacky way to get around the token limit and add new knowledge to it we're not training it and then cooking the information at the Base level so having a a chat history on chat CPT is fantastic so you need to get as close as possible on this side of things as well and so by
now you may be wondering what's the difference between this and chatbase I'll show you if you give me a second so first of all you can do the prompt editing same as with the chat base I think this is actually a paid feature within Dante AI so you will have to upgrade to their basic plan for 9 or 20 a month or something to get Access to it but I can do the same thing here I drop in the prompt for my Jerry's gym AI so I can put that in there and then I can
go next step and then you depending on the level of your plan you have access to all of these cool functions there's a ton of really cool customization features here firstly I can edit the suggested messages here so I can add another one saying how can I sign up for PT classes so we can have suggested questions down here in the Bottom right so that as soon as they hop onto the chatbot they can click one of them it's a lot more user friendly to help them find frequently asked questions you can remove the powered
by Dante depending on your plan you can remove this microphone Etc so there's a whole bunch of things custom styling custom fonts even I really like the editor and it allows you to make a really customized experience when you send it over to your client as A prototype there is also the opportunity to do custom URLs depending on your plan I believe you have to be on the 300 a month so it's definitely very up there and probably not something you want to do initially but being able to put this onto a custom URL and
instead of having the danta AI a URL you can actually put a client's custom domain on it and be able to have the sort of white labeled and customized a chatbot at least for the client to use so that's a Big step up and I haven't seen many of these prototyping kind of softwares allow you to put it out on a custom domain like that so one of the features I really like about Dante AI so if I confirm all those edits I now get a direct link as I can see here if I post
this over I can have a chat to my chat bot here what are your membership plans Etc this is the kind of thing you send over to the client more importantly though you can go back and you have an Iframe that they can put it onto their website so they can embed it into a one of their web pages then you also have the chatbot bubble the same as we had previously with chatbase you can copy this and then put it onto their web page as I just did before and it will pop up as
a little bubble down the corner so all your favorite kind of deployment options with one of these prototyping tools but one of the key things about danta AI which I really like is that It's mobile responsive when you're on their web page so in this web page version the web chat if I inspect it this is just explaining it how it works on the phone but basically when you actually have this web page up on your mobile it's going to to look normal the proportions are going to be right and you're going to be able
to do all the things that you'd expect to be able to do on it just through the web browser on your mobile phone so this is a great Prototyping platform with a lot of flexibility a lot of customization in terms of The Styling and the property Etc so um really like this and as you can see it is a little bit different from chat based pricing is different but I thought I'd run you through this one as well just to give you an idea of what it does and just to give you a quick look
at the pricing on this they have some really good lower tier options at 15 30 which Tend to do a lot of the different functions that you need so have a play around see if you like it and if not there's another one coming so we can check that out now and finally the last prototyping platform I want to cover is called Kodi AI this is a little bit different to the other two as they have been this one gives you slightly different use cases and different deployment options so this can be very good for
some cases and I just really Want you guys to have a full view of the different prototyping options you have and how they can fit into dealing with different clients as always you can sign up to Cody in the description below I do have an affiliate link for that one so if you want to support me please use it once you're signed up for an account with Cody you're able to come onto this dashboard Here and Now quick orientation we have the chat we have Bots content and account where do you want to start In
terms of creating these customized chatbot prototypes is to go to content now in this situation you can see I've created three different folders one called client One customer support one called client one marketing my uncle client one reporting now this is the real point of difference for these chat bots made by Kodi AI is that you can create these different folders as you saw earlier in the video and by creating different folders it makes it much more Easy to not only manage the different information that you want to have access to on different chat Bots
but also update them over time so they've got a new reporting document for a client I have a chatbot that's put onto their slack and it allows their managers to chat to it and get reporting information whenever a new week or month finishes up and we have a new report I can easily come on here and throw the report in there and then without the client having To do anything it's just going to update their knowledge base and they're going to be able to continue to chat with it as if it just happened magically so
I really like that the folders featured here it can be very good so in this example my client one could be in e-commerce store and we have a folder for the customer support related documents a folder for their marketing related document in a folder for their reporting related documents and so I can Create a chat bot as you're going to see in a second to connect to the customer support documents and then I could use that to put onto a web chat pop up so that people can ask questions on their website Etc so that's
going to handle their customer support and then when it comes to internal reporting I can connect that to the slack Channel as I just mentioned they can have a chat away to it and they can continue to upload documents and then for anything related To marketing I can have its own sort of Silo of information that I can add to over time and this is a great way to keep things segmented and be able to update them in real time as your client needs them so to demo this we can open up the customer support
document we've got my jiri's gym document in there already you can just upload it here it takes websites you can drop a link on it as well they don't have the video URL option like Dante AI does but you can Still import our website and strip the sitemap you can upload a document as I've done here and then we want to go back to bots and I want to I'll just delete this one for now then we can go up to a new bot on the corner and then it gives you a bunch of cool
for you guys who are not too familiar with prompting and things like this there are some great presets here so if you want to click on a factual one more of a document question And answer if you want to click on uh customer support they have a preset for that now what these are all doing is as I assume under the hood each of these different buttons has a specific prompt that when you click on it it's going to use that prompt in your chatbot so not rocket science they'll probably just tested a whole bunch
of different options until they've got a a couple that work for different use cases now because we're super smart and we've done All our chatbot Theory just an earlier in this video we can actually go to the advanced mode yes I know what I'm doing and now we get control over the prompt we get control over the system prompt as well this relevant score as well so there's quite a lot of features under the hood for Kodi AI which is a really surprising for a platform like this and then there's even token distribution I love
this so you can change how much context how much history and really have A lot of files that you can play around with so I'd say Cody AI in terms of a prototype platform is really pushing the limits in terms of how far these things can go before you have to get into the more advanced options but they've done a really good job to show a whole bunch a whole wide range of options and settings you can change and for someone like me this is heaven because there's so many different ways that I can tweak
it and options for me is the bot developer to Try to fix it if it's not working the way the client wants in this case I'm going to leave all of this the same apart from the prompt as you've seen before I'm going to be using this prompt that I've used in the previous sections so we've got my name is Dave an AI assistant tasked with answering questions about Jerry's jump then we can go next bot name we can call this Dave from Jerry's gym and do the same there and so here's the different models
you Can select if you get the premium one you have gpt4 in this case we just need GPT 3.5 which will do fine and then here's the key bit so I can easily select which folder I want to get the knowledge base from so in this case this is James gem Customer Support I'm going to click on the customer support checkbox but if I was making Jerry's gym reporting I can just click on the reporting option there easily create a chat bot and then deploy it to wherever We want to do it so I can
create this now I can chat to it by clicking this instantly up open hi Dave so yeah they've done what are your locations in the east coast and boom Boston it's giving the links as well so it's using the prompt where I've told it to deliver at markdown yeah this is great and I've really copied the chair TPT UI here but the more important part for us is not necessarily using it within this but we can share it we can Turn this on you guys are going to see my link but I can copy this
and have a chat over here I can show the client this I can send this to them for testing they can do the same thing hi Dave what is your support email so this is the web chat deployment option that they give you I know it doesn't look great like this but there are a ton of different customization features that you can do so feel free to play around with that you can also get it on one of these Bubbles and just like the other prototyping options we have a shareable link that takes you to
the full web page we have an option to embed it on the site as an iframe so it just blocks into their existing website then we also have the pop-up embed that you can add to a little bubble in the corner you do have to add their domain if you want to actually test it so be sure to get the domain and put it in here it's a little bit of info here in terms of how can Make wild cards so it shows up on every page Etc but this is a little bit of setup
you can do and that gives you a clear idea of how you can use Kodi you can get different folders and create different knowledge spaces for a client you can manage it all you could be managing 10 20 different clients within this content The Hub really and you can set it up exactly how I have have client One customer support client one marketing client one reporting create All the different Bots and the cool thing about Kodi is that once you've built all these different knowledge bases out and created the Bots you can super easily connect
them to Integrations as you go to account and Integrations here you have slack Discord and jpo this AP is actually coming soon but slack is really big one if you're going to be working with companies like marketing agencies and in any company really that's using site for Communication you can create a chat bot you can add it to a slack Channel and make it available to chat for all of the people in the workspace really cool deployments and I think you guys are going to think up a ton of cool options for this but this
is another prototyping platform that I like to use as you can see it's different to the other ones each of them is good for their own little thing but that about wraps it up for prototyping or you can actually get Done with the basic stuff and start getting into the fun stuff which is actually building powerful robots with voice flow now the first thing you need to do in order to follow along with the next part of this tutorial is to sign up to voice flow which is going to be the chatbot builder of choice
for the rest of the tutorial it is a very powerful platform and there's a number of reasons why we're using it mainly because I think if you guys are going to be Learning a new skill and really starting to become a chatbot developer you need to be starting off on something that has enough depth where you can build a really expert skill set now saying you're a chatbot developer on Kodi AI it's there's no depth to the platform the reason why I'm really liking voice flow at the moment is because there is so much depth
to it I've been attending to the team and they've shown me how that massive and how complex you can Make the applications here they have a ton of Enterprise clients so like large businesses BMW JPMorgan those big companies are using the exact same bot building software that we're going to be using today so I figured if I'm going to start you off on a proper bot Builder why not choose the one that gives you the most sort of depth that you can learn and really become an expert on because if you're going to follow the
skill and follow it see where you can go With it this is going to be your new home so we may as well start to get your board up to speed on how to use it and the different functions and features it has if you are going to sign up and follow along I'd appreciate it if you could use my link in the description it helps me to keep these resources free for you guys I continue to make these monster videos so that I can keep doing what I love doing also before you dive head
first and follow me into signing up For this platform the pricing is also very affordable as well it's going to be 50 a month for the Pro Plan which has a ton of great features but you do have the sandbox option to follow along with me within this video so they do have a free trial I definitely sign up for that and give a get a proper taste of all the different options they have but then beyond that it's also a shared editor so as you start to build a team if you're going to collaborate
with other people Hey figma style editor so you can get multiple people on there working leaving comments doing all sorts of stuff so I really like using this with my team at my automation agency and another reason why I think this is very fairly priced is that it combines a lot of the features that I've been using with other softwares into one so you basically have all of your body needs knowledge bases and everything for what fifty dollars a month on the Pro Plan so definitely Check this out to see 100 clear those are the
pricing options now we can get stuck into using this thing once you're inside voice flow we have a couple things uh just for the home screen here first of all you have your team obviously I'm with my Morningside AI team so you can add different workspaces you've got a resource Hub a education hub for them which is great so if you want to really take a deep dive into this platform and come up to speed even More after this video definitely jump into this I've got a ton of resources uh essential tutorials here they've got
a YouTube channel as well so if you really want to get ahead feel free to dive into that back on the dashboard here they have templates which are really cool as well so if you want to come in here browse around the different templates they have it might save you a bunch of time for a future project and they have what's new the Discord Community is Great as well if you want to sign up to that and then you have basically the the invite section so you can invite different collaborators so if you have a
team member you want to bring on you can invite them here and then you have the important button which is to add a new assistant which we're going to be clicking now so in this case we're going to be doing it's called customer support we can click build in your assistant continue select the channel we're going To be choosing web chat for this one as you can see there's different deployment options so you've been learning all the stuff like that we've got the web chat deployment we've got twilio SMS deployment we've got WhatsApp we've got
Microsoft teams one thing that I'm really looking forward to them dropping is this Facebook Messenger deployment because that's something I like to do with my lead generation chat Bots but here you can go web chat for now select The language English and continue so once you've got that you can create the assistant and then you're going to see a screen that looks a little bit like this so this here is the voice flow editor it looks nice and simple which is what you want but it does have a lot of features packed into it just
walking you through firstly we have the designer tab on the left which is what we end right now so this is called the designer then we have the all-important knowledge base section So we're going to be adding some stuff into this shortly but you can add data sources up here select the different types of data all sorts of things then we have transcripts which you're not going to need to worry about just now here's a little transcript from previously you have analytics which is a very important thing to have Within These chat Bots how you
have the Integrations tab which is going to allow you to export the the functionality to a Or query your knowledge base and all sorts of things I think this is Advanced and we're going to be covering this probably in a later video but there's some really cool functionality that voice file allows you to do out of the box and then finally you have your project settings which is just some basic stuff the name Global no matches and your navigation methods I'm using this on a trackpad I need to probably change this over to trackpad now
and Then there's a couple different settings you can play around with this feel free to get comfortable down here we have the number of tokens you've used on your AI usage so they have been limit that you're allowed to use up in terms of AI usage throughout your different applications I've obviously got a quite a bit of token usage left you guys will probably have a smaller amount based on the trial that you've started but we can start jumping through the different Features here up in the top left we have the sort of home menu
just going to allow you to do a whole bunch of important admin things with the chat bot you can export it you can clone it you can go back to your dashboard Etc we have different Mouse types as we've just been playing around with you can leave a note so I can drop it on here and say something a nice way to just annotate it for your other team members I can drop an image right a video in here all sorts Of different things in commenting of course as I mentioned before if I want to
leave a comment I can jump here and say my teammate's done something stupid I can tell him that he's stupid right here and then if I want to go into the nlu model this is just basically a high level look at all of the different intents and things you've got set up so we'll be probably using this a little bit later in the video and we're doing the advanced builds but for now we won't Need to look into that too much on the left panel here we have topics and our topics are essentially going to
be Pages let's call them layouts or canvases maybe this is the home topic but I can create a new one and I call it topic two and then I can start an entirely new sort of page of my application but I can still access it through the chatbots sort of interface and through the flow of the conversation so if I wanted to build something out within the home that Links to an entirely new page maybe it's like an entirely different section of the chat bot that's going to handle a very different bunch of features whenever
the user hits a certain trigger in the home topic we can shoot them over to this new topic it might be the refunds topic or something like this and we've got an entire page built out it's just mainly for sorting the different parts of your application in order to keep it a bit clearer and not having it All modeled on the one page we won't be needing this for now let's just close us down fit back to the home one and we can delete this and that's most of the orientation we of course have the
publish button up here which we'll be using later and then we have the Run button so whenever you want to test the functionality as we're going to be doing soon you can just hit the r key or the run it allows you to run a test of the bot and then we have the share button so If you want to send the Prototype over to a client in the same way that we've done with the prototyping softwares this of course has the ability to send over a quick prototype so we're going to be seeing this
very soon you can change the branding the appearance as I've done here invite your teammates export and all sorts of stuff and up here you can easily change the name of your chat bot to whatever you want just edit it here I don't need to do mine for now and I will Mention one more thing and that voice flow has their own assistant down here it's actually built on voice flow I've had a look at it it is insane the amount of complexity these guys have built into their own assistant using their own technology if
you need any help this thing is going to be very good at helping you out so if you've got any questions if you don't know what something does if you're running into an error I really urge you to give this Thing a go and see how much you can get done if you get stuck so this guy will probably end up being your best friend as you're learning this platform now with that helps keeping out of the way we can jump straight into sort of checking out these features now I've gone over this section A
couple times and I've tried to walk you through all of these one by one and show you what it does but really without context they don't mean anything so the best thing we Can do is just jump straight into building our first chat bot and you can learn on the go all of the concepts and things we talked about earlier in the video are going to come into play in full force as we're building this try to connect up the dots as to the things I've mentioned and how they look in real chat Bots but
with that out of the way we can jump straight into building our basic customer support chat bot with voice flow now my use of the term basic Throughout this video is going to refer to something that's a pure AI chatbot as I mentioned earlier in the video we have pure AI chat Bots old school chat Bots and we have these modern ones and in this case they're going to be custom knowledge chatbots so I say a sort of infinite loop with a prompted AI plus a knowledge base so we can get stuck into creating new
knowledge base for this and and setting up this Loop that's going to go around and around now when you're Approaching building a chatbot for a business it is important to understand what your goal is and what you're actually trying to achieve not only for you as a developer to know when you've got it right but to clearly communicate to the customer what to expect so you'll find a lot of the time if you don't get those the balance right between expectations and what you're actually able to deliver you're going to have issues with clients it's
important to Say this chatbot is a basic customer support chatbot it is intended to deal with 80 of the common queries that we're going to get in this case we're going to be talking about an e-commerce store called Pet Supply or PetSmart or something whatever I've called it all of the information that I've included in the knowledge base document is meant to handle the 80 of questions so it's not meant to be a Beal and indoor solution it's meant to say I don't know and send It off to a customer support email when it's a
more complex question they can't answer so the 80 is still going to be a great help to the e-commerce business and we're going to work to build this first at 80 rate and then when we go to the advanced one we're going to try and really push the boundaries and and set something up that's going to best basically eliminate the entire customer support team now the most important thing to get right for one of these to Work how you expect is obviously the knowledge base itself so when you're making a custom knowledge chatbot like this
that relies so heavily on a knowledge base it's important to make sure the knowledge base content and the document is actually good in high quality it has all the information you need because that's going to really sting you in the butt if you don't have the right information available so here is an example of a e-commerce store Called petmates that I got chat CPT to generate um it's not very good this is probably not a great example so here's one for a Jerry's gym one the example we used before now this is a lot closer
to what you actually want to be working with there's a bunch of background information about the business we've got the contact information section which has got the support email we've got the gym locations the billing inquiries app support all of the locations and Addresses Etc so this is really in-depth offering trials FAQs as much information as you can get and throughout it we are dropping the links to the support email so if this happens then send a support email we also have links to the parts of the website when it answers a question that's related
to a membership system it's going to send them a link to Jerry's gym prices so having the links included into the document for these kind of chat Bots where they're dealing With 80 that's ideal for you to have the links included in the resource document so that it's able to send those during the chat as well and really direct the people to the correct resources since it doesn't have that full range of functionality to to deal with everything themselves but because the advanced version of the customer support chatbot that we're going to build after this
is using the petmates example I'm going to be using petmates as our knowledge base For this so we are here on our home page we can head over to the knowledge base section list these resources and these documents are going to be available in the resource Hub so as I said anything that prompts the documents as well going to be available to download so you can follow along step by step with this link is going to be in the description first step is to go to add data source we're going to add a doc I'm
going to click on the Petmate stock load that in there and It's going to load for a few seconds shouldn't take too long so that's all processed and ready to go now we need to go to the settings section and so in here as you can see I have the Dave prompt what I want to do is actually spin this to be for petmates.com or whatever website we're talking about so you can just go to chat GPT and take this exact same prompt that you guys are going to be able to get which works for
customer support in most cases drop it In here give a bit more information about the new company we want and it's going to shuffle it around and make it about the company that you're looking to make it for change this prompt to be for a e-commerce pet product called pet that's the name of the assistant is Annabelle and there we go it's writing this out it's a great way to just repurpose The Prompt for whatever different business you're doing so feel free to take this prompt send it to Chatgpt and say hey look this is
the business I'm working with now can you re-shuffle this thing around to work for the business somewhere so we can copy all of this and head back the exact version of this is going to be available on the resource Hub so you don't necessarily need to follow along here but we can paste this in I want the max tokens to be about 200 temperature can be low I'm going to be using Claude V1 for this which I've been recommended as Is pretty good so we can use Claude V1 in this case scroll down and save
it and then when it comes to your knowledge base it's important to check that it's giving you the right responses so what we can do is once this thing goes away we can hit the preview button and then we can ask it a question so we go what information do we have in our resource Hub okay we can say how can I return an item to initiate a return log into your pipmates account go to order history and Share return login to Facebook yep so that's responding correctly and it's telling us the right things to
do so we've got the knowledge base set up we've tested it to make sure it's working then we can head back to the designer and start to actually set up the conversation design for this so step one is going to be grabbing a text and dragging it out here now we can put in here hi my name is Annabelle from pitt put a little dog emoji and there's a Cool feature here where you can generate a couple different variations so it's not always going to be the same there we go hello Amanda Bell from hip
Bates greetings I'm Annabelle from Peter mates and you can hit enter on each of these or you can just hit command enter to save them all looks good to me then we need another talk step so we're going to put another text one here you can just zoom in on this say how can I help you today okay we're back for day two of Filming if you couldn't tell I was getting a bit dark so we're back the next day got a new cup of coffee ready to go so let's jump straight back into it
but while I have a moment could you please if you're enjoying the video if you've got anything out of it so far please head down and leave a like on it would mean the world to me it helps me reach more people it helps me continue to do what I do on YouTube so if you're enjoyed the video please leave a like if You like this kind of content and you're interested in seeing more please subscribe and hit the Bell if you want to be notified but in the next one drops I'm also going to
be doing another one of my 20 minute Consulting call giveaways so if you want to go on the draw to win a Consulting call with me have a chat to me about your business about chatbot building whatever you want to talk about I'm all yours go down below and leave a comment on the video Telling me what you like about chat by building what you're interested in doing with it Etc or the hardest thing that you've struggled with within this whole process so I want to know more about you guys and what you're struggling with
it really helps me as a Creator so please hit down below there'll be some instructions in the pin comment on how to get into in the jaw to win that but best of luck and I'm going to be drawing One winner every month so this is going To be an ongoing giveaway if you're watching this video 10 months in the future you're still going to be able to get access to going into the draw to win that so hit down below leave a comment and I'll be uh replying to you and then you can get
in contact with me on Discord and I'll be adding a name to the pin comment as well best of luck and let's get back to the video so back we left off we have our first text card my name is Annabelle from pepmates and then We have the follow-up message as well so we can connect this up here and then if you want to quickly write and test your app you can just click R and we can hit run test and then there we go conversation has ended great now next thing is going to be
hit escape and we can draw up here we can take a listen step so this is going to be capturing some information we want to capture something that the user says after that so hello how can I help you Want to catch that information now we get into the first interaction with variables within voice flow so this is a variable called last utterance we have the last utterance which is typically the last message that the user has sent and then we have the last response which is the most recent response that the AI has sent
so good thing to distinguish early on in this case we're going to be saving what the user says to us as the last utterance we're going to be Capturing the entire user reply so the entire message they send and this is a little bit more complicated we will cover entities later but then this here is where we're going to be saving the response so we're going to take the entire user apply if we're going to save it as the last utterance now we can tape it to a bunch of different variables in here I can
create a new variable and I can save it to that but in this case last utterance is just fine for us we're Going to wait for the user to message us and then we're going to a that reply for later so next step is to get a response AI box these blocks enable you to send a message back to the user based on some prompting whether you're using an AI model sort of out of the box or using your knowledge base in this case we're going to be using our knowledge base and then we're going
to be saying answer the user's question and then what we can do is reference the last utterance so this Is going to sort of insert what the last article tool saved at and then it's going to inject it in there so if I go preview and I say what is your refund policy it's going to inject that into here answer the user's question what is your last refund what is your refund policy and there we go we're getting a preview back making sure this is all working and then the little magic here is that we
can just Loop this back up to the top here so understanding how these Little loops work is very important so as you can see here we've just got a little Loop going where we ask them how they're going and then we start the conversation they ask us a question we save it to the last utterance we pass that to the AI to answer from a knowledge base and then we just pass it straight back up and open the floor again so let's see this in action if I click here and go R run test hello
Amanda Bell from petmates how can I help you what is your refund policy and then it's going to take that question save it as a last I don't want to take it down to the AI and then insert it into the prompt that we have here and you can also click play from here which is helpful if we go what is your refund policy and then it's just taking it straight back up to the top and we can ask another question so it's just looping around now understanding these Loops is really important because We're going
to be using them a lot and this is how you can keep the AI kind of in an endless conversation there's a couple cool features we can build off the back of it but that's the basics right there and then we can ask another question how can I return my so how can I return an item to initiate a return Etc what about refunds but then it's also working so in this case I've talked to the voice photo team and they said that there is chat history here so as we Mentioned earlier in the video
in the concept section chat history is important especially for these kind of conversational chat Bots where you're working back with a customer support agent so the chat history is a thing here I will say it's not amazing it's not the best it's not super strong like chat gbt is but it is enough to have a little bit of context I definitely think there's room for improvement on this so I'll be talking to the team at voice Flow and seeing if there's anything that they can do to potentially Crank that up and make it a lot
more Chanty like when it comes to chat history so this is the basic Loop for a customer support chat but now in this case we don't have many links and things being put in here but the knowledge base itself so you're working with a different business you need to include lots and lots of links to the different parts of the website so say they ask about pricing or they ask To manage a membership you can just be dropping directly from the knowledge base as your response you say hey we can manage your prices here or
you can change the system about your membership here's where you can do it and you can provide the link and so they're going to click through to that and get all the information they need or you can just start to drop the support email if there's any questions that are not within the scope of the chatbot but this Is a pure AI chat bot just set up on a loop that's going to continually answer questions so this is just the start and what we're going to be doing now is going to an advanced customer support
chat bot which really expands all this functionality lets voice flow really Flex its muscles and terms of what it can do for these kind of chatbots so we're now going from a basic customer support chat bot over to the advanced side of things now the first thing we're Going to do is actually delete this connection here then we're going to select these select this and we can just Chuck that down there somewhere and what we're going to do now is drag this out and get a new block which is going to be called a choice
block now this is a very very interesting and important thing within voice flow what we're going to do is set up a number of different intents so that the chatbot can pick up on intent from the user rather than just Answering sort of from the knowledge base the whole time so we're going to be building in a number of functions into this chatbot as we've mentioned earlier in the video to make it do cool stuff and not just be stuck in this endless loop of being a pure AI chatbot so we're trying to identify that
the user wants a specific thing in this case we're going to have things like they might want to leave a review on our e-commerce store they might want to leave some feedback Etc so we're going to build in a couple different features and then if none of those intents are detected in the user's message when we go around this Loop then we're going to just reply with the knowledge base is normal so we're going to pick off a couple different cool things and then have some sort of fallback which is just okay we'll just respond
with the knowledge base here we have the choices what we're going to have one is a choice a choice a choice Actually we can just click that so we just need three paths here so the first intent or the first function we're going to add to this is tracking that the user might want to track their order so we can come here and click on this we need to create a new intent so this is the how you set up an intent within voice flow so we need to give a couple examples to the system
to the AI to the natural language understanding system that's underpinning all of this give Them a few examples of the kind of things that might trigger this intent well in this case we're going to be saying a track order and then we're going to say where is my package that's how can I see where my package is can I get an update on my pack where is my order so I'm giving these manual examples just to really fill this out but what you can then do once you've got a couple good ones I really like
what you want to do I like to click give five New utterances so this is going to use AI to generate more like it how long will it take for my order to arrive what is the status of my order can I track the shipment in my order yep these will seem good so you can either go through and just press enter on each of these one at a time like that or if you want to do will generate even more make sure that they're still on par as you can see this utterance level the
strength by here the more examples we give it the Stronger this is going to be and the better it's going to be at picking this up than the chat so what you can also do I will mention is if all of these look good what you can do is just hit command and return and it's going to approve all of them so this is going to save you a bit of time that I've found now we have our track order intent so this is going to pick up any time the user is asking about where
their order is and then we can do special things with it so we can Create that intent there I'll just leave that there for now then we're going to create our next one we're just going to be leave a review I'm so in this case we're looking for things like I love my product my audit arrives so fast I am very happy with my purchase I highly recommend your score so these kind of things where it's just raving about the product recommending the product Etc we want to be able to catch this and and Get
them to write a review on the product essentially so I was very satisfied with my purchase I'm impressed by the service here so pleased with what I bought definitely we'll be back to purchase more telling my friends so we can just go to command and enter on this I do quite like all of those that's good let's leave a review we can create that intent and then I'm going to create one more this one is going to be something about looking for products so looking For that in this case because we're in online pet store
we're a pet smart or whatever break all this we're going to be putting in some keywords for what people might be looking for in terms of a product so what toys do you have for large dogs um for cats what products do you have do you stock any Dogs so we can see what they want to generate us here if they gotten the right idea what type of products do you have do you have any dog food um do you offer any dog food products which products do you carry okay that looks good to me
maybe we can throw in some for cats as well um what cat food Do you have and now we can create the intent so now we have our three different intents so this is going to try and capture it as the message comes in same as before we're going to ask them how they can help but then we're going to potentially pass it through the intent classifier and give it a couple different routes that it can go before it ends up at the fall back so before back in this case we can set up here
which is going to be in The settings we can have a node match and we can add it manually just going to leave this empty I'm going to create a new path and then what we can do is just route this over to here then we can take the no match and just route it over here for now and do the same sort of response AI step and then here we're going to have this exact same thing as we had on the earlier setup we had on the basic chatbot which is just to answer the
user's question I'm gonna head back over here add it in as the prompt select knowledge base and paste that in answer the user's question now we have all of the different intents set up in terms of the different examples and it's ready to identify this message as it comes in now we need to actually set up what happens so to do this we can set up an intent here drop this and what we want to do first let's start with lever review so when they come down here and we detect that they Are wanting to
leave a review potentially based on them loving the product then we want to have an action we can go to intent then we can go to leave a review this is essentially going to teleport us up to this part of the chat so we're ready to start leaving a review so if they reach this part of the conversation up here they've probably just said based on our intent something good about the product so I really loved the experience the shipping time is so Fast so we can jump straight into sort of going for the throat
here and asking for a review so text you can say that's great to hear because I've just said something nice about the business and then we can drop another text card so it's not too sort of Blocky then we can say would you be interested in leaving a review I can send you a discount code for your next order then what we want to do is use a new feature that we haven't touched on yet which is buttons so we Can add a button here that says yes sure and then we can add another button
here by clicking add button and make it safe no thanks and then if they say yeah sure you can come up here and start a new block add a new text card so they've let us know that they're interested in leaving a review for a discount so we can say amazing can you please provide the product name and something you liked about the product and your experience shopping with us with this I can create A little message for your review so we're going to use an AI feature just here to generate them a review based
on like one sentence from them so it really expands on the review and then all they need to do is approve it so we have that put in there then we can add a capture step in because we want to hear what they say with this I can leave a little message for your review so we're going to capture the entire user reply which is going to include a name of the Product and something they liked about the service and then we're going to take this up here and we're going to use what's called a
set AI step so we've used the response AI before which is going to generate a a response back to the user what we want to do is just sort of the set AI can work without necessarily putting something out in the conversation so it's basically like running a little chat CPT function you might send a bit of information to it Tell it to do something with it and then you can save that information back as a variable and use that variable later on so as you'll see if we use the set AI card here we
can say generate review so this is just the name of the particular card here and we're not going to be using our knowledge base this time and what we're going to be doing is just processing the information that we got from the user which is being captured in the last utterance variable and then we Want to prompt this AI into generating us a little message for them to send back so in this case I have the prompt ready to go now if you see me go a little bit fast do not worry because all of
these templates and everything I'm building here is going to be available on my resource Hub so at any point you want to copy some of this and not have to write it out by hand you can head down below and grab the template then you can import it into voice flow and You have all of the stuff I've been working on all at your fingertips and you can even just take that and start using it so if you're struggling at all if I'm going too fast there is going to be the template in the description
for you to go over yourself at your own pace and when you pick apart what I'm doing here and why it works how it works so and really make it your own as well if you want to modify it feel free to do that and start selling it if you really Want to so the prompt that we're giving this is that the following is information from a customer on the product you purchased from our eCommerce store pet mates and what you like about the product or experience don't mention e-commerce please create two to three sentence
of review based on this information that is cheerful and reflects well on the business and then all you need to do is put a colon and then we're going to reference the last Utterance that we've received here so we asked the user for some information we capture it into the last items variable and then we just inject it into this prompt here so what we want to do is create a new variable we've already made it like what you can do is just click the create new variable and name it the exact same thing so
we're going to be calling it tracking API response and then in the prompt settings we're all good maybe we can turn this temperature Down a bit I don't want it to be too too creative here and then we have the max tokens we can probably bring that down a bit and yeah now we can do a preview so if I hit this we can actually test what it would output if we put something in so I could say I bought the doggy bottle chain really fast it's super easy to use so I like this with
a voicemail how you can very very easily test the outputs I absolutely love the portable dog you bought it from petmates I've only did it Arrive quickly but it's also incredibly user friendly I highly recommend this product and this wonderful company I mean it's a little bit overboard you can prompt this more as you want but I think for the purpose of this tutorial this is going to be fine now we have our output from the review generator then we need to actually save the output of this to a variable so we can use it
later in our application in this case I've already created it called review but you can Just click here and create a new variable and create one called review also you can click here and it's going to save our output to their variable and then we can add a text card and we can get the ok so we're going to say thanks is this okay to post for you and then we can provide the review this is just going to be again inserting the variable into the chat so this is how the Syntax for it is
going to be using the curly braces around it in this case we're Using the curly braces around the review object and I put some quotations as well to let them know that we're kind of quoting what the review will be eventually then to approve it we just need to add a button here we can add looks great here I want to add another button also and we say no I don't like that okay so now we have these two outputs we're going to tell them where they want to go we're going to have a talk
step here and we can have another Talk step here this is an interesting chance to take a look at the kind of design elements that make the the editor voice Globe feel a lot less boring and plain so we can have a red block for the no route and a green block for the yes route so makes it a bit clearer without having to name all these different blocks we can see very clearly what this sort of happy route or unhappy router so it looks great or no I don't like it so to edit the
colors you just need to Right click on it and it's going to show you what color and you can use the slider Etc you can add comments you can save the library all sorts of things if the user approves it we can say amazing I'll post this on your behalf now then if you're actually building this up with a really common store you could either integrate it into their system so they'll automatically or you can use API or something like this to basically take this information send it to a Google Sheet for a VA to
go through and start to add these reviews and manually once a week or once a day whatever you want to do and then if they say no they don't like it we can say I'm sorry I'll try harder next time please leave a review on our product page I'm not going to bother within this application but we can Loop it back around and try to revise it and have sort of a back and forth with them until they're happy with it but for the sake of this tutorial I Think this is good enough this is
the branch of leaving a review ending a helpful little node to add a sort of a reset I found as you put a text on just here above what I'm going to call our main Loop so this is our main Loop here what we're going to do is when we can send this one back here on its own but right here we can just make the chatbot say how is there anything else I can help you with maybe generate a couple different versions of it just something Else I can do for you sure that looks
good that's good to name this one back to me look maybe I'll give this a color and then what we need to do is connect the back to main Loop up to the main Loop of course and then what we can do here is a cool feature of voice floaters to be able to drag this and drop it here and we can actually go to actions go to block so here we can actually type in back to main Loop and now instead of having to draw this big Arrow out that Goes across the whole screen
we could just make this little thing here that's going to shoot them back over to this and start it again so I think with that we're good to actually test this out we can click the Run button here but you can hit run test now because we've played around without intents every time you add new things to intents new utterances and things like this you actually need to do what voice locals training I'm not 100 sure what they Actually do under the hood I'm sure it's just bundling up the intense and putting that into a
system-wide prompt or something anyway you need to click this button and it trains it I know we're supposed to not say that but they do something here figure it out anyway you just need to give it a few seconds to dial in the natural language understanding system and then we're ready to go so hi we can just restart this actually just click on the plus Button here click on the play button on start hello I'm available for petmates how can I help so the first thing I want to test is that this no match is
working so that if none of the intents are triggered we're still sending it to sort of a regular customer support knowledge base so if I ask something about nothing to do with ordering nothing to do with reviewing nothing to do with what products we're looking for if I just say what is your refund policy it should Route it all the way down to the bottom here and then answer it from the knowledge base yep and our return policy is great so that side of things is working now we need to test if it's sent us
back up to the top of the main Loop now we can ask you I'm really happy with my purchase boom and then what we've done is it's taken us up to the leave review section as we planned because it picked up on our intent that we were happy with the product and now it's Taken us to this section up here sort of teleported us up that's great to hear would you be interested in leaving a review I continue with discount code for your next order and then we go yes sure amazing can you please provide
a product name Etc I've just run into a bit of an interesting issue here guys so it's important that you see this stuff live so what we've set up is an intent that listens for when they say they've loved the product right so if I say that's Great would you be interested in leaving a review yes sure and I go I loved my doggy bottle okay so so fast and my dog loves it let's copy that now if I send that we think it's going to go onto this generate review section but what it's actually
doing is jumping back to here here you can see it's important to come up here and click the debug mode actually so if you want to turn that on you can see a lot more but we're jumping back to this legal review section now What's actually happening is it's picking up on the intent we created originally to look so what we need to do to fix this and to make sure it goes through to this generate review block is to click on this come to the settings here and then when it comes to intent scoping
only intense in this step so that's going to prevent it from listening to Global intents and it's going to run through as we expect so if we go here leave a review hit the play Button yes sure I'd like to leave a review I've also set up the no route here you can see I've just taken node to go back to the main Loop and it asks hey how can I help you with anything else now we can paste this in here I love my doggy wallet came so fast and the dog loves it now
it's going to go to this step yeah there we go thanks is this okay for you to post the bottle is perfect for our walks and hikers so convenient looks great I'll post this For you and it shoots us back is there anything else I can help you with and bang we're back to the main Loop so that is one example of how you can set up an intent and build a cool little function around it I'm going to start to speed things up now that you understand how intense work how a lot of these
cards work what these AI things the AI features do and how we're kind of routing things around this app in general or going back to this main Loop A lot of the time before sending them back to this so this again where we are right now in the conversation is back on this main Loop and it's ready to pick up any intent from our questions or it's ready to send us to the knowledge base again if we say what is your refund policy it's going to send us down there okay now I'm going to speed
run through one more of these intents that I think I really want to show you guys how to use external API calls to get information I Will send information at least first things first we need to go to the event section click on intent drag it down here so we have a new little area I'm going to call this track and Order and then we want to have this go to track an order so if the intent is triggered we can go to track an order now I don't want this particular function to do is
actually the text first of all via intent classification if they're asking for a an update on their order or They're wondering where their Packages Etc so that's going to trigger this coming out here then we're going to ask them for the order number we're going to check our our database and pull some information back on that order number and then use AI to manipulate the response and give them a natural language response to where the package is so first text is going to be please provide us with your order number so this is obviously coming
off the back of Something mentioning about like where's my package because that is what was set up in the utterances for this particular intent and sorry guys if you're not sure what utterances are then each one of these different lines or ways that you can trigger this intent is called an utterance so we have intent which is the whole thing and then we have utterances which are examples of showing that kind of intent so just to clear that up there first first message here is going to be Please provide your order number and I'll check
the status for you then you want to open the floor to the user to message us so it's going to be a capture step we're listening for a response we're going to capture the entire user reply now in this case we could actually set up a custom entity to try to extract the the order number but it doesn't come sort of out of the box so for now we're not going to touch intense we're just going to do pass the exact order number And assume they're giving it to us in the right format and then
what we're going to do is come over here send another talk step and let them know that we're checking in on their order and now for the real fun stuff this is what I really like to do with these kind of Bot Builders is to actually integrate it or make it access some external information and pull it into the chat and make it more contextual make it more valuable all that sort of stuff so in this case What you would be doing is taking that order number and then depending on your e-commerce client querying yeah
after ship API or whatever wherever they're handling the track and maybe even the Shopify API if you want to be pulling directly from the store you'll probably be using a tracking API like aftership in this case I don't have any Commerce clients to do this on currently so I'm going to be setting up a a zapier store just to put a little bit of dummy data In and simulate how we might pull some information back in this case we're going to be making a get request so we're going to be asking to get information from
my Xavier store and that's going to be done with this URL here now this goes back to our API section we'll be talking about requests and responses in this case we're going to request information on a specific order from our API so to get this figured out you're probably going to Need to look at the documentation for the particular API that you want to use in this case I've retested it all out I can just paste it my JPS store URL here so that's basically the number that I'm going to be calling to get the
resources that I want within this is going to be as you can see over here on Postman I've done a little bit of testing this is the information that's going to be returned when we call it so I have the order number and then I have some tracking Updates I will just generated as such as dummy data but ideally you'd be getting this from after ship or something like this and it's pulling back all of the latest tracking information what we want to do is is grab this whole thing based on the order number and
then we're going to pass that to the AI to write a response to it so we have the URL we have it set to get we don't need anything in the headers we don't need anything in the parameters now I can't Leave my zapier store exposed for all of you to to hop on and start pinging so I'm not going to be able to show you the exact URL but this is just a placeholder and a dummy for what you would eventually do with the clients so in this case I'm going to create a new
variable it's going to be called tracking API response you can create it here I've already got it what we're going to do is send a request we can just click the send request button here And it's going to we've got a 500 error so internal server error oh yeah we got that error because I haven't put the S colon slash slash and we can send the response and there we go we've got all of our information back so this is being stored if you just want to see if your store is a very basic way
of storing a bit of information absolutely just a web page with this this information as I've done it so you can add a ton here I've just got the one but in this case we can Pull this information back from the store and then we can pass it over to an exponent of our applications we have the information we're saving it to the variable of tracking API response and then we're going to be taking that onto the next step so if the API call succeeds we're going to do something up here actually it's going to
be a JavaScript step and if it fails we're going to do something down here and we can say a text step again maybe I'll Change this block color to be red and I'll change this to B Green okay now in order to do something useful with the information we've gotten back from the API what we need to do is first do a little bit of code and that was scary but don't worry what we're doing is just calling the variable that we've just saved and we're taking all of this this code let's let's call it
code it's called Json and what we're doing is is stringifying it and turning it from this Sort of object structure which is a a data type just like sort of flattening it down simplifying it down to just being a string so we're just making it into something that we can pass around within our application and use is within our AI query so we're just using json.stringify you're going to be able to get this code from from the templates it's very very basic so I'm sure you guys can probably just read it from there and copy
it and then what we need To do is take that information and create a response out of it so what we're going to do is use the AI step here we're going to use a set AI or let's pop here and then what we're going to be doing is using the AI model this is going to be summarize tracking update so we've got a little prompt here please give us a brief status report to the customers asking about the order here is the latest tracking info and then what we're doing is pasting in all of
the Information that's just come from the API that we've recorded take made the API call saved it to a variable we've flattened that down into a string that we can use within this part of our application and then we're using some example responses saying given all of this information these are the kind of example responses we want and then also reminding that you're responding mid conversation it's not do you need an introduction just just give us the Message and that's we need to do so we want to apply this to I think I've already got
it saved here order up update so you can create a new variable and just call it order update I'm going to save it there and then we can actually preview this so I'm going to preview it get the information from here this is the example that's going to come back from the API now we can generate and see what it's going to give us your order number one two three four has been Shipped by China Post and has arrived at the original Depot of Los Angeles Etc so this is a good little update on the
order that we can send back to them we've done a good job to take the information from the API process it and I'll send it back so we have saved this to our order update variable now we can just go here and go order up there great so now we can actually just test this quickly so there's one little thing I missed out here that I need to add this Response word into here so make sure you've added into the capture response you're also capturing What's called the response a good way to debug this if
you're having issues is to actually just sometimes it'll be returning stuff but not actually saving it or doing something wrong it's actually a good idea to add a text card here and actually just put the response so I was able to see it printed out in the chat what was going wrong so if I go up here And I can play from this part again let's go to prints out what's actually being returned within the chat here so there we go I can see the information we're getting and then this is the response it's getting
so now I know it's working I can just delete this come here click that right click and delete and then we're good to go there we go so now we're getting a nice little response back and then after that we can just send them back to go to Block back to my look and if this doesn't work issue with Qing order information then we can just go and get actions go to block back to main Loop sweet so that's about as much as I want to do within this application for you all so let's run
through it quickly we can we don't have this look for product section that's sort of just dummy dummy information there now you can very easily add that in and have some kind of Assistant that helps them find the chatbot you can have a knowledge base that is all the information on your products and they can ask back and forth and help them find products so I'm not going to cover that within this video it will be here all day we can run through here now so if I say where is my package jumps up here
great please provide the order number okay my order number is one two three four great now I'm going to check for information thanks checking Out this won't be shown in the actual chat it's just the the debug mode shows you all the information so we can turn this off up here and it won't appear so your order's been shipped it's excellent is there something else I can do for you and so that's a quick walkthrough of an advanced or modern chatbot that uses an intent classifier Loop so every time we come back down to this
we're going to run through it ask them they can ask a question ideally we can pick up on a Specific thing that they're interested in doing and then we have built out specific functions to deal with that that interact with the business that pull the data from different places they can do valuable things Beyond just talking to them in this little bubble so what we can actually do now is publish our chatbot I can show you this we can go publish we'll call it a V1 publish it so it's not successfully published we can embed
the widget so we can click on This and we have all of this information here so we can go to the web chat section it's going to actually help you do it here but we're smart boys and girls so we know what to do so you can play around with the settings here you can play around the appearance customization I'll let you guys do all the fun stuff there but there's a lot of things you can do in order to customize how it appears and I can copy this and do as I did before just
give you a Little example of what this looks like I can come in here and edit to there boot up the site so when it comes to actually putting it on your client site you can just give them this code and they'll be able to add it on themselves as we can see down here in the corner we have the started web chat hey my name is Annabelle from petmaids how can I haven't customized this as you can tell but this is what your client will be able to have on their website so we can
Remove The Branding and things like this where is my order please write order one two three four and there we go back to what we're just doing so this is how you can put it onto a website you can provide the client or whoever you're working with with this particular little script you can make all the customizations you want and that's how you can actually deploy to a web chat there's also another cool thing I want to show you briefly when that is if you Want to share it we can go share the Prototype so
we're actually just working on this as a as a prototype and you want to show the client you can open this link in a new tab and here we have a actually password protected if you want to do it so you can add a password in the effort called Jerry's and you can go Jerry's view the Prototype and there we go we're into the chat you can send this to clients to demo with I love live and there you go so that's how to make the Beginnings or release the foundations of an advanced customer support
chat bot using intent you can prototype it you can send it over to a client with this you can send them the script tags out onto their website all sorts of options here but that is the web chat deployment free a advanced customer Sports outbot now what we're going to be doing is jumping into a very very similar example but we're going to be deploying it to WhatsApp as well so if we go back to the Dashboard how to get started with our WhatsApp staff training chatbot we can go to the new assistant we can
type in staff training chat what we can go build an assistant and go next and then when you select the channel you need to select WhatsApp and language English whatever you want to do and this is going to ensure that we're setting it up on the WhatsApp Channel Once you click create and when you've done that you should see a blank canvas like this with The WhatsApp similar at the top knowing that you've in a WhatsApp deployment style chat but I've already got my one here so the point of a staff training chatbot is to
allow access to all of the relevant information and training documentation within their business making it available through a sort of permission list and around the clock access to this information so whenever a employee or a contract or something has a question about the business instead of Having to call up a manager and start off a whole communication string of emails or text or whatever it is basically distracting the people higher up in the business from the important tasks you need to be doing instead they're dealing with little small Castle over here and they're trying to
answer questions for people who are in that ranks below them so by centralizing all of this information into a custom models chatbot you can take it and give it Access to them in this case we're going to be doing it through a WhatsApp chat bot so that whenever say you have a contracting business and someone has an issue with something on site or they can just pull up the WhatsApp chat and actually ask a question say hey what are you supposed to do when this happens or who's the person to contact if there's a leak
in this building or something so all of this information can be put into a chatbot it can get everyone on the Same page and this can be a very cool thing to offer in terms of your automation package in terms of this and this can be a cool deliverable and particularly handy in specific Industries so it's not going to be amazing for all Industries but there's some specific areas where these kind of things can be very very valuable so I thought I'd include it in this video to show you another application for this chatbot technology
and voicemail in Particular okay so we're going to be starting with the basic customer support chatbot and it actually uses is the exact same loop as we've just done previously which is a short little introduction about who the person is you're talking to and then it goes into this main AI Loop so this is the pure AI chatbot kind of style and this is going to be answering from a knowledge base that we've created in this case I have it over here training.box I can pull This up to give you a little bit of an
idea again I've just created this with chat gbt but in this case you'd want to get as much information as you can on the business that you're working with and really a comprehensive guide on on how to be an employee at that company in this particular case we're talking about a solar company called Smith solar and this particular chatbot is a sales rep training chat bot so when you're on board a new sales rep this is going to Have all of the information in it so that they want to know okay what are the what's
the voltage or what are the Watts of our particular solar panel what are the brand that we work with what are our payment terms are there any special discounts that we can offer at the moment so this kind of information can make it really accessible to these new sales reps and you bring them up to speed and even with common objections and things like this so This has a lot Of information I have fairly big discounts referral discounts payment plan options common objections and solutions Etc CRM and Lead processing so what the sales rep should
actually be doing he or she can just say hey I've just closed the client what am I supposed to do in the CRM and it's going to tell them the exact steps you need to take so you get the team all working on the exact same information and then clearly go on doing the things you want Them to do at the right time so this document is of course the probably the biggest and most important part spending your time making sure this document as A1 is going to be a much more valuable than spending a
ton of time building the chatbot out so this basic Loop will do the job for you in most cases but we're going to add a little bit of functionality just to make it a bit cooler so you've got your document again this is going to be available on my Resource Hub if you want to get out and follow along I've added my data source here uploaded it and then what we have is the same loop as before you can go back to the project we just did and copy and paste this in and I've changed
this around to being a guy called Angelo from Smith solar sales rep assistant I'm here to help with any questions you have what would you like to know so I can just connect this up here intro message a main link and now getting a major part Of any of these customers chatbots when it looks so simple like this there's usually a lot of complexity happening elsewhere and in this case it's happening on a training document and how that's structured and the information that is included in it but then also we have the prompt as well
so we have the settings up here in the knowledge base maybe I'll change this over to quad V1 for now temperature I'm going to have it nice and Low Max tokens and then I've Got this prompt here this is going to be available in the resource Hub so if you want to make a a sort of sales rep stuff assistant chat bot you can use this exact same prompt as well and then we're going to be able to save that come back and give this thing a spin so we can see play from here what
discounts can offer and there we go here are the discounts we can offer referral discounts early grade discounts payment plans Etc what Is a PPA PPA is a power purchase agreement so it's all working good we've been able to ask it questions escapes get out of there the cool thing about this is that we can actually test it directly on WhatsApp send us out as demos to clients when you set up your app for the first time it should have asked you a bit of information about where you might be interacting with this on WhatsApp
so I've already given it my number so this option pops up here on The top right and I can click test on WhatsApp and just like that that messages me sent to my number and I can update it up here if I want to change it to a client's number but if I go over to my WhatsApp here now this can work on my phone or on my laptop I'm just doing it here for ease of use it's popped up YT staff training is really fused okay launch the assistant then it's going to kick the
chat off what would you like to know I'm Angela what how do I use our CRM all right what is the best way to handle the cost objection they are saying it's too expensive and then we go we're able to chat to the chatbot with on WhatsApp now this kind of WhatsApp deployment is good for testing and sending to clients for demos to approve when it comes to actually deploying it really for production you're going to want to get your client to set up a twilio account and follow the steps here available on The documentation
for voice flow so that's a bit different to actually get it up and running in a in a proper production sense but it's fairly easy to set up with things like twilio you're going to need to be using your clients twilio account for this so just be sure to communicate that early on that this needs to be set up on theirs and at the end you're going to be able to connect the the app that you've made to your WhatsApp business account that they're Using and then you're going to be able to deploy that and
have this thing ready for production and ready for General use by the team now in order to take this to an advanced chat but we're going to follow the same kind of process but we're going to take an attempt classifier and we're going to add a function in this case I'm just going to do the one because I think you get the message but we're going to be doing something a little bit cooler with what We're doing within that specific action so we can start to make the same modifications as we did last time which
is to disconnect this we can select this by holding shift and then selecting it all we can move this guy away a little bit and then we need to get an intent classifier going so what would you like to know we can go listen and we want to go choice you have the special action or function that I want to build into this app and it's going to be a incident Reporting so this can be good for all sorts of sort of instruction or any any labor kind of businesses where there might be incidents to
report this can be very very important for health and safety purposes and things like this so I want to have a function where it's able to detect if they want to report an incident and then take them to a specific set of steps and then we're going to do something with that incident data and which is really cool I'm going To show you in a second so first off we need to actually set up the intent for reporting an answer so we need to click on the choice block we've added here head over to create
a new intent I'm going to call this a report incident and you can say all right report into the event report and accident reports and issue there's been an answer I need to report an incident Etc so I've got a couple here now I can Just generate 10 more see what they give us there's been an accident yep that checks out reporting issue report next and I want to report a problem register an issue document an accident file a report file up it's not a complaint I need to report an incident sure okay so we've
got our tent set up with a very strong excellent level of utterances you can create that intent and then we need to have the fallback option which can be set up here no match we can add manually And then we're going to add a path to this as well so what we can do with the path is just send it over here what we can do actually save a little bit of time let's just head back over here and grab this and we can drop this over here we can delete these so if there's no
match we can send it to the knowledge base answer the user's question excellent then we're going to Route this back around maybe put it down here we need to delete this so if it does to Get that report incident intent then what we can do is add it up here we're going to go report an incident I hope you guys are starting to see the pattern here now and then what we can do is take them through okay what's first thing when you say okay first thing you might ask is is this urgent is there
an emergency or something like this we can say is there an emergency okay is there an emergency now as you can see on the listen side of things there aren't Actually buttons with any basic deployment of WhatsApp within voice flow you don't have access to buttons but you can do it on the advanced setup there's a little bit of custom coding you need to do with a riplet so we don't actually have that option here but we can still do a choice and we can use natural language input to determine what they want so path
one we can have this as a yes option and then are we going to add another path as well and these are built Out of the box so as you can see we're just selecting an intent voice loads obviously got a pretty good idea of what a no intent might be so this isn't going to necessarily give them a button it's just going to open the floor for us to say what we want so if I just run it from here actually let's just set this up this is the emergency fix okay so if it's
an emergency please call 9-1-1 and get the help you need from the Agency Services we don't really want the chat away and what's happening is emergency so they should be getting straight to the action but if it's not an emergency then we can say okay can you provide some info on what happened include the following rocks time of incident where the incident occurred three who was fall and what happened okay so we're asking them for a bit of information about the incident with this info I can Okay we can just test this out make sure
our intent is working correctly and all the stuff up here is as well what would you like to know report and we need to train the assistant like that when you add a new intent utterances you need to train it again as I mentioned earlier we just need to do one little thing here click on this we can do actions and we can go to intent and in this case we're going to be going of course two deep report Incident That's in so we can play This again run test report and incident is it an
emergency there we go no it's not great so that's all working fine we can play this again from here with the play button again and because we've set up an intent it doesn't need to be a strict Max so I can say sure rather than yes so that's because uh The Voice flow intent has a bunch of different options that we built into it and the AI has a good understanding of what is a sort of yes intent or a no intent so we've got Our non-button using routing here so we can just provide natural
language in let's go into relative to the different places now we need to take in this input click the capture it's going to take in the information we're going to to save the entire user applied to the last utterance yep that's all good with me next we're going to use a set AI step to extract a bunch of information out of what they've just said so they can provide a natural language so they can Even like use dictation and just say hey I was standing outside on the porch I'm working on a project and then
Dan came along and this happened and he got an Island's foot and this is the issue so they can provide that all in sort of a jumbled mess of ticks but we're going to use the AI to do is extract that information and put it into variables and then we can use this variable for the cool next step that's coming up in a second extract entities okay so I've Done a little bit of prompting here I'm looking to firstly extract the date and time so you are an expert text extract that your job is to
identify information within a body of text and pluck out the specific information for the entity you're looking for now actually I might just make this a system-wide prompt cut this out so we can go to the prompt settings and add this as a system message so all of these are going to be text extractors or information Extractors we need to turn the temperature way down we don't need that many tokens at all we don't overspend and then what we're going to do is I've told it from the text provider to extract the date or and
or time and provide it in the format day month month year Etc use 24-hour time assume that the year is 2023 but you're going to save this to a new variable called date and time we're going to create a variable and then we're going to do the Same thing but for the different entities so for the text provided extract any any names I mentioned we're going to add this to a new variable called names and then we're going to do it again add another one and then we have from the text related extract any locations
building names or place names we're going to call this locations create new variable and what else do we have what happened contact provided extract effect what the topic Of the text is and identify what the incident was and what happened then we can call what happened we can create a new variable put the temperature way down system prompt or ready to go you're an expert text extractor your drivers too so we just need to add this little bit at the bottom to each of them so insert the variable come down come down pasting these into
all of them and then what we can do is actually preview it so if we Click back on here preview now we can say at 12 M on the 12th apart if I was at the on and leave with very big and he slept on a first pass wood on the deck that was delivery and Earth is heading back okay we can generate see if we get the outputs we want okay we've got a couple tweaks to make here I wasn't too happy with it I mean what you need to start doing here is providing
examples so say you put an import and output input and output then you can really Dial in exactly how you want it to respond for the purposes of this tutorial I'm not going to play around with it too much but I think for now this will do the job if I preview this it's going to take us through all these different things okay so it's got the date and time coming out names mentioned Barry b yeah there's a little bit more prompting that can get rid of this stuff so we just get the action variable
we have the locations building names Placement names topic incident what happened Okay so we've got the information out now we've got them saved to the variables now for the cool part which is actually to take this information and push it over to a Google sheet so that we can analyze or do something with all of this information weekly so we want to analyze all the incidents that happened in the past week okay where are they where are they happening most often we can look through The spreadsheet Etc so easy data collection on the employee side
of things and then it can be easy analysis of all these different things on the sort of management side and admin sites in order to get this information out of our chatbot over to a Google sheet we need to start to use automation platforms which is one of the things I wanted to show you guys how to do within chatbots so right now we're just going to quickly go through using one of the Automation platforms that you can use and show you how to connect your chatbot to at least a Google sheet but the same
principles applies where you can send the information over and then manipulator on the automation side of things so I'm not going to be covering the full in-depth of the automation platforms I'm actually saving that for a future video so if you want to know more about the sort of other side of this business model where you can build the Chat Bots but also build the end immigrations and how it actually interacts with the business be sure to subscribe to the channel because I have a another big video coming on that to show you guys everything
you need to know there now first things first we need to decide on what platform we're going to be using in order to create our automation so in this case we have zapier which is something that I've used quite a lot but there's also make.com Which is something I've been getting recommended a lot by my community members just purely on the cost basis so it costs eight times less I think to use make it is a little bit more complicated but I believe it has a lot more depth in this tutorial I'm actually going to
be signing up myself and trying out make I haven't used it much before so you're kind of going to get to see me learning live alongside you but we're going to set up I'm sure it does the exact same Functionality as zapier we're going to be using an automation with the mate since you we can send the information up to make it's going to handle it we can tell it what to do with it and add it to a Google sheet so first things first we need to actually create an account on make link to
sign up to make is going to be in the description I think I do have an affiliate link for that one as well so again if you want to support me in the channel you can use that link if not The domain is making.com it's not very hard to remember so once you go over to make.com we're going to sign up or sign in going to log in here once you're in the make dashboard here we can head over to templates now I think we need to go to web hook we can click on web
hook and then hit the first one add data to Google sheet receive via web hook you can click on this they've got a template ready for us to use so we can click Start guided setup what you're going to Do is create a new hook we can click add I'm going to call this incident reporting web hook I can save this now it gives us this URL here which we can copy to our clipboard and then we can head back over to our voice flow and now we need to drag out an API section very
similar to what we did in the last one but we need to put a get request here I believe we're doing then we need to set this as a post request because we're going to be posting some information Over to make.com and then we need to add in our parameters the information that we want to send has been all saved to variables from this extract entity step that we just did so that was the point of extracting it all into different variables now we can go into the body and we can start to go what
did we get out of this we have the date and time the name change the location and what happens so we can go date and then we can go date then we can add another one And we can go names we can go people we can go names then we can add another one we can go not people but they locations we'll go location and then we can go finally we can go what happened what happened and now we have all of that information packed up we can send a request off to make we can
hit the send request button let's go to uh names Liam location Germany what happened got rained on for three weeks which is what it's been doing here I don't know what's Up with the German weather but send the request and we've got a 200 back which means it should have gone through now we can head back over to make.com successfully determined great that means it's received our information we can go continue now we need to create the Google sheet that's going to click this information so we can go over to sheeter we're going to create
a new sheet and we can call this incident reporting and then we can do the same thing date so These are going to correspond to what you've set up on the voiceplay side of things so we have the date the names location what happened so now we have all of that we can head back over to this you need to connect your Google account it'll prompt you to do that then you click on the spreadsheet ID and we can go to or is it incident reporting which has popped up here all nice sheet name just
sheet1 is fine then we can go continue Then you can click on this again and now we can start to actually assign the values here so the information we sent from voice flow is going to appear here so you can see all the different ones that we've ascent and the body so if we go back to this all of these bits of data that we sent as the body of our post request has now popped up here for us to populate the different fields with within our spreadsheet so we have the date we can line
up with that we have The names you can put as people I've had a bit of a mismatch here a location nice location of course and this one I take it is going to be what happened that hasn't come through yet but yeah that's all set up and we can actually start to test this thing so we can turn the schedule on and now we should be able to Ping this thing again head back send a request go to Liam home fun send request okay and so to make sure this is running non-stop in the
background you need to Turn your scheduling on so you need to flip the switch here and the settings should be immediately as the data arrives then you can head back to the home screen and then we can see here if we pop back over to our request we can hit resend and it's going to pipe play a new one here and we can resend it's going to populate a new one here so this means every time we run a new request through this it's going to process it it's going to add up to the sheet
and Then we can say a little bit of information back to the user so if it succeeds we can say text thanks for submitting report in a fail so we can say fail so please try again okay so now we can actually give this thing a spin maybe we should do it on WhatsApp as well let's try to do that so I can send a test on WhatsApp report and incident okay is there an emergency no there's not oh this isn't formatted very well and WhatsApp might need to tweak that But last on the 10th
of August James and I were walking and the car of C block and the floor broke which made us fall down and badly scrape out legs so we've given the information now it's going to process that and hopefully we see oh there we go we've got the information that's been put over to the sheet now again as I said this needs to be a little bit more prompting done to make sure this is extracted because all you need is the names thing so you need to Do a bit of a few shot prompting it's probably
going to do the job here also for the date and struggling to get the date so a little bit more prompting and information can help there did we get the C Block C block yeah topic incident that happened what happened James and the and the writer we're walking Corona C block at the floor break causing them to okay so now we built an incident reporting feature into our chat bot we were able to take some information from The user process it and extract some entities albeit badly at this stage and I submit there is as
I said there's many improvements you can make that extraction but for the sake of this video I think it shows you the concept and you can feel free to make the edits you want on your own time but but we're able to use make.com to use a web hook and send data using a API call and we're able to send information to make.com using a post request with the variables That we extracted and then it's going to put that into a spreadsheet for us to analyze at a later date now this is just one example
you can of course use this template of using make.com and sending information over and creating a spreadsheet for it and things like this so many different things you could do with once the data is inmate.com I know there's a lot more flexibility there than what zapier at least so you can manipulate the data however you want you Can send it to different parts of the business but when it comes to the automation site using platforms like make that deserves its own video so at some point very soon I'll be releasing that video on it as
well so if you want to cover this side of the information in terms of chatbot building and then also the automation side be sure to subscribe to the channel because that's coming very soon okay so with that feature added and I think we're pretty happy With pulling an end to the staff training chat bot for now we took it from a basic chatbot with just a loop were you able to ask questions about the knowledge base and now now we've taken it to having an intent classifier and building in just one feature of many that
you could potentially build out this is again just to show you the potential of this kind of technology and we've used for the first time connecting to an automation platform via a API call Or a get request post request and we've sent information out of the chat box so that we can manipulate and do cool things with it elsewhere and we've shown you how to do a quick and dirty deployment to WhatsApp so that you can test it within a live environment you can send this out to your client and get it tested and whatnot
so happy with this as it currently stands again you can get this template available on the resource Hub which is going to be available in The description so feel free to grab this download it import it to voiceplay I think is an import feature in the top left on my reasons Hub when you download this it's also going to be a little video on how you can upload this so once you download the file for my resource Hub you need to upload it so you can start using it so there'll be a video on the
resource hub for that but for now let's get into our final chatbot which is on lead generation okay now we can Get onto our final chat bot which is going to be a lead generation chat bot I wanted to include this one to give you all a very broad range of different applications you can use chatbots for so first off we have customer support which can handle a ton of different use cases then we have a staff training chatbot which is a good example of the point to WhatsApp and then finally we have this lead
generation chatbot which really can provide the most value to a business in A lot of cases if you set it up right and now it's important to understand the concept behind this and why we're doing it and how we're actually going to generate leads with a chatbot like this so the idea behind this is we're going to create a chatbot that has specific knowledge within a specific industry and with that specific knowledge we're going to put it into a chat bot and make it freely available to people be it through our website chat bot or
on maybe a Facebook messenger chat bot or even Instagram DM chat bot for a business there's so many different places that you can put it but the idea is that you have some interesting information that's specific to a user's problem and you're going to allow them to get access to that information and basically speak to an expert on the topic and they can ask the questions back and forth hey maybe they're looking to buy something maybe they want to know more information about A certain area you're going to give them access to that but in
exchange for giving that free information away you are going to be dropping ad placements and in this case it's going to be lead captures throughout the conversation every three to five messages we're going to be dropping our own lead capture method so that when they're mid conversation we can be pushing our own products and really the key thing here is that if someone is interacting with Your specific chatbot for the business that you're working with they are obviously very very interested in what they have to sell or the information regarding the niche they are a
warm audience they have shown interest and intent to get information and potentially be a customer to some kind of service provider in the niche so your goal is to get the people on here get them chatting back and forth with it and then start to push them and say hey look By the way we're actually experts and we can help you with this or we have this resource that you can get click through here to get it or you can book in a call with us and have a chat to us Etc so that's the
concept in this case we're going to be doing it on solar I've created with chat gpt's help and ebook resource I think this is a good way to go about it if you create some kind of comprehensive ebook or specific knowledge and put it into a document and In this way we're going to chop it all up and put it into our knowledge base and that's going to cover all the bases and people are going to be able to ask questions about the solar ebook that I've created and the solar e book is specifically for
single-family homeowners who are interested in getting solar for their house the idea is that this Smith solar fantasy company that I've created has this chat by putting their website or maybe on their Instagram or something like this and when people come on they can chat to it and they're going to learn about how they can get solar for their home and it's specifically for residential solar and they get to ask questions about financing about the Best Brands about the best installers and things like this and every couple of messages we're going to be shooting in
our own lead captures and things like that so that's what we're doing here you can see firstly the Knowledge base of course is going to be the most important part of this at the beginning here is the knowledge base here it is about a massive ebook a decent sized thing that I've gone back and forth for chat gbt to generate and this includes things like why consider solar energy for your home evaluating your solar potential basically all the things that someone in this Niche could ask about we're going to have information in the symbiota provide
Value to them so it's important to have the knowledge base as a sort of comprehensive package of information that is going to be specific and useful to the people who are going to be using it so to get started we need to head over to our newest button we can type in lead gen and we're going to build an AI assistant of course once we're inside first things first we need to set up our knowledge base I've already got things set up here so I'll just delete that Same sort of process what we're going to
do is go add data source we're going to click on the doc we're going to select our document for these examples we're just using one but you can of course add a ton of information in here unless it just for Simplicity of the tutorial if you're doing the staff training chat what you can just cram that full of Staff training resources for this case if you want a broad knowledge base on a specific Niche you can start it full of Any documents that your client has you do need to do a little bit of prepping
though to make sure that it hasn't got a bunch of rubbish in it but the idea is you can put a ton of different ones in here so in this case I'm going to put my solar document in give that a second to load in and then we can go to our settings up here as always I'm going to be using Claude V1 here turn the temperature down a little bit Max tokens can probably stay around there and then We have the system prompt which is very similar to the other ones this is for a
guy called Nick who's our solar energy assistant AI your mission is to help a single family home owners understand everything about solar energy enabling them to make the best decisions for their specific needs my name is mentioned things about the context Snippets that you're going to be provided with Etc and instances where the query Falls Outside of your information or ability you should guide homeowners to seek advice from a qualified professional in such cases direct them to Smith solar speak to an expert page with a clickable link using the markdown format like this so this
is going to be providing links to Smith Solas speak to an expert page I try to make these Legion chat Bots kind of unaffiliated if possible because there's nothing that puts a prospect off like instantly bragging about how your Businesses so in this case you can tell I've said you're just a solar energy assistant AI you're kind of just helping them your goal is to just help them get the information they need not to be pushing products really hard but then when it comes to speaking to an expert you can gently sort of nudge them
towards it because we don't need the Chapel to be selling selling selling because eventually we're going to have our own ad placements that are being put In there and those will do the selling for us so again this prompt is going to be available on my resource Hub if you haven't got it and also the entire template for this chatbot going forward so we can save that we've got a knowledge base set up and then we can head back so this looks very similar to the the prior ones we've done just a little introduction hi
I'm Nick I'm an expert on solo for single-family homes what would you like to know and we're Going to go through the same sort of process which is to capture the user input listen for capture and then we're going to catch the entire user reply in the last utterance variable and then we want to do a response AI block pop that here connect this down here then we want to use the knowledge base and we say of course transfer users question let me go last utterance so we can preview this if we go what is
the Best way to finance solar based on the information provided this is a common issue that you'll see with custom knowledge chatbots which is they mentioned that you're providing them with information this is an important thing for us to go over actually I'm glad that this has come up so I'd say based on the information provided there are a few good options for financing your solar system so we don't really want this AI to be saying hey I've been Given a bunch of information we just need it to speak as if it knows it like
that's what it actually knows so you can prompt the living hell out of this thing to try to get it to stop doing this I believe I've already put it in twice you guys never refer to these Snippets in your answers all the info given you should be repeated as your own knowledge and then I do it at the end okay so I've done a little bit of trimming on the on the knowledge base prompt here I've Added in well you must never refer to these Snippets in your answer I've also said but you must
never refer to this information directly in your answer so maybe we can save this and try it again it might actually be a voice flow related thing whether they're injecting a little bit of prompting into it themselves when they're running their own knowledge base so let's just give this a spell and Steve's question preview what's the best way to find it So long let's see what it says okay it's still doing it we might need to get a little bit more strict on this guy this is really strange if I can put it at the
bottom here you must never say things like based on the you provided or similar never refer to the N4 formation right like you know all information yourself so this is actually some good stuff for you guys to be seeing behind the scenes because you're going to run into this a Lot where it's not quite giving you the responses you want let's try it again there we go the best way to find it solo and it just goes straight into it so there's a little trick that you can do with prompting and sort of for every
Cutting Edge stuff that we've been working with the Morningside and reading it out of some scientific papers but there's a thing called getting lost in the middle when it comes to prompting where depending on the size of your Prompt if it gets really really big the stuff in the middle actually gets kind of lost and the instructions in the middle aren't noticed as much as the ones at the stop at the start and at the end so as you can see I added some stuff sort of in the middle there and it wasn't listening but
if you slam it at the bottom it's definitely going to take a lot more notice of it and that's a general rule with prompting as well is that the longer the prompt get the more Bloated it is the more fluff you have within your prompt the less notice it's going to take at the specific point so if possible make it direct make it short my one's very very long here I could definitely do some work to tidy that up but it is quite a long process it's literally like writing like copywriting or doing a really
really specific writing task and being very very precise with your words for the sake of spinning output things like this I find this Prompt does a good job but when it comes to a specific client and you really want to dial this in you should go back to the drawing board and plug things out and very very carefully lay out your goals for the chat bot and how you can instruct it on those things very very carefully and keeping in mind that the very first thing you mentioned in the prompt and the very last thing
are going to be taking the most notice of and the stuff in between may not be looked at as Much that's a little bit of extra sauce for you guys when you're doing your prompting as we can see that's solved the issue for us here so we've got the setup now we just need to reconnect this Loop up to the top and then we have a bit of a prototype let's run this run this test hi I'm Nick I'm an expert for solar and single family homes how can I help you I'm interested in solar
but I'm not sure about if it's worth it in the long run solar energy is absolutely Worth in the long run for most homeowners here are some key reasons why environmental benefits energy cost savings Etc so this is quite a long answer and as you can see it's actually been cut off due to the Token limit within the knowledge base but this does look like a lot because we're doing it within the little skinny editor but if you put this out onto the prototyping page it would probably fill out a lot more and look a
little bit better you Can add something at the Ender at the top in order to prompt this thing into putting out shorter responses so that's all good there we've got a little prototype running maybe ask it one more thing we say what is a PPA a power purchase agreement or PPA is a financing option for residential solar panels okay great now you may be thinking this is just a custom models chat bot what's the difference between this and what we've just done in a real business if you're Working with a real client and you wanted
to prototype this what you can do is include links into the knowledge base a lot more often because this is a dummy business we don't have the links to put in but when you're working with a real client you can include links to the the checkout page or the book A call Page Etc all the kind of products that they have available and things like this or even just a link to the page on about us and about our services and why we're Experts or something or you can even be doing blog posts and things
like that if you want to learn more about this specific side of solar for residential is a great article that we've found that kind of warms them up to the idea of working with some solo so in this case when you're just running this basic one and we haven't got the sort of manual ad placements that we're going to set up in the advanced version and your property you can say something like you're Supposed to help them to learn more about solar for residential homes your end goal is to help them come to a decision
about whether they do or don't and each of your answers were possible try to point the user towards Smith solar as being the ideal provider for their needs that they can really help them to get what they want then as they go through the conversation they ask about who's a good provider they can say hey by the way Smith soul is pretty good Here's the link to book and a call or hey I'm interested in figuring out more about installation costs and say well if you want to learn more about installation costs so a solo
actually Stacks up very very well against the competitors we have very competitive pricings on this system this if you want to book in a call or if you want to schedule a consultation here's how you can do it okay with the basic one out of the way we can jump into the advanced One now the first thing we need to do is set up some variables you can click on the nlu model at the top here here you have drop downs for different entities Etc what we want to do is add some variables in you
can click the plus button here what we want is to have is just sent we can create the variable as you can see I've already made it but I won't have that error for you and then we also want to have router same sort of capitalization you can have lowercase on The first word and then each additional word is capitals after that at the first letter we can create this as you'll see I've already created it so you've got these two new variables what you want to do is when we start off the chat bot
here we can put a set step in and we want to set up variables apply to the add router I want to set the add router to B1 and we're going to add another set I want to set the message of sent to zero not always necessary to do this but I do like to just make sure everything's set up at the start of every chatbot correctly okay and the next step is to grab this JavaScript block and put it here what we're going to do is go messages sent and then just type space plus
plus this is going to add one message to the message Ascent variable so this is just a counter variable that's going to calculate the number of messages that we've sent it's going to keep track every time we send a message It's going to add one more and then we need one more block just to test this thing out so we can set a conditional if block here which is going to be check for trigger number of messages so this is going to say if the variable which is the message sent if the number I've missed
send is three then we're going to want to do something and then we're going to have an else which is a no match so we can add a path if this is not true we're going to go back up here And then we're going to come over here if it is true so we can add a torque step I'm just going to say add triggered okay so you may be wondering what the hell's going on here we've set up these two variables of adrollter and the mess descent add realtor is going to be the variable
that manages what ads people see and when and any number of messages sent is just to check how many messages have come through so that when it hits a certain number in this case we're going To make it every three messages then we're going to push and add placement in and then continue the conversation so we can run this and it should have the debug window on the side here you can pull this out let me just get this thing out and then we're going to be able to see our variables on the side here
which is very handy okay so when we run this thing we can watch our messages sent on the side panel here look hi Nick and we're going to see our missed descent Increase by one don't let this put you off you haven't actually set up any buttons this is actually just a setting in the guided navigation so if you turn this off and you won't actually see those things so this is going to stop and wait and give you buttons on the earth step so you can really control it so if you run this again
say hi Nick and you see our messages sent will increase by one now the ad router is still set at one and we can go what is your name my Name is Nick I'm an expert in single family homes what is solar so the ad was triggered which is because you sent the message we've also sent ours and so we sent ours it added one to the message here and then it sent his reply and then I put it through this Earth conditional block again and because we'd hit three messages it sent us over to
the add browser so this is all working as expected so now we have the opportunity to start to place ads in front of the User as they're going through so now we're going to build out all of the content for that side of things so what we're going to do is have three different ad placements I'm going to plot them all out now each of them are going to start with a little JavaScript so we can pop this JavaScript one here oh I can install this so we need to go messages sent equals zero then
we need to go add router equals in this case it's going to be two don't worry too Much I'll explain this in a second paste this again paste this again rename this to add one add two add three so we're clear and now this one needs to be the add router set to two and this one needs to set the ad router to three and this one is to set the ad relative back to one so this is just the corresponding ad so after we see the first ad it's going to update the ad relative
variable so it's going to be relative to the second one so it's basically just cycling Through every time it shows an ad it's going to change the ad that's going to be shown next a little bit more Logic on that is coming soon okay to start off the first ad which is going to be placing an ebook in front of them that they can download and in exchange for the email or their phone number so we're going to have three different ads in this case I'll show you the complete version over here which is going
to be one an ebook that they can click through To and claim secondly it's going to take them to a solar savings calculator which is going to have a lead capture on it as well and then finally we have add three which is just going to be a traditional lead cap so it's going to ask for the email or their phone number then we're going to take that phone number and email and put it onto a Google sheet so that we can track them later and do things with them so those are each of the
ads we need to add in our card first There we go this title is going to be claim your free ebook we're going to browse this I have a little image made for this that I whipped up residential solar so you can add all your own images in here I'm going to provide these all on the resource Hub as I've mentioned so if you want to get any of these and follow along you can I'm going to add a description to this as well so a little line under it so this is going to pop
up as a cool little card within the chatbot And then we can go a button we can say claim yours so they can claim it for free is the idea and then we want to open a URL I'm going to do my Liam otsley.com then we're going to add in another card down here as we have done before what else can I help you with generate let's get a couple different versions of that anything else I can help you with yeah sure that sounds good yep generate more and then what we're going to do is
write from this link down Here over to this so this is going to be the home base we can go this back to main loop as we've done before maybe give this a nice little color because I'm feeling happy and then we can take this and send it back to our main loop again so same sort of stuff we've done before just this time with some cool extra features in it we can pop this up here now we need to do ad number two which is also a card so I can add a card in
here drop it below I can add in the Title I can add in the subtitle and then what I can do is browse this I have it on my desktop here just a little picture of the calculator that they can use and then I'm going to add a button as well it's going to go to try free it'll be a good call to action here and then we're going to open a URL we have a URL here so this is a solar calculator by Google of course you'll be using one that your client has it's
going to allow them to calculate something or some kind Of tool that's going to capture the lead as well so we have that there and then we can link this bad boy back up here same as before and then we need to do our last ad placement which is going to be a little bit different what we're going to do is just add in a text block after this JavaScript block here and we can paste this in and then this is the message we want to send would you like one of our experts to get
in touch with you if so what is your preferred contact Method and then what we can do is have some buttons pop them down here and we can have that email and we can add another button and say phone now you can also set up an intent here so that they can type in just natural language rather than a button so you can add an intent here which is emailed as fine or do a bunch of different utterances on that so that it can identify that they're asking for email same thing with phone like you
can create your custom intent here I Think create a new intent do the same process that we've done before so that they can enter both by button and they can also enter by those typing in what they want to do as well so if you've got both of these set up and they're going to come up here the email we can put talk step phone we can do the same thing text please provide a valid email and maybe generate three different versions of that there we go please provide a valid phone number generate one of
those Just adding a little bit of variation so that if people are going through this multiple times it doesn't seem too repetitive no harm really they make it super easy just to button click and that's it okay so I've gone ahead a little bit just to make sure everything's planned out carefully for you guys one thing we do need to add in here is to add another button saying I'm okay thanks so this is basically the option for them to not give their Information so we're throwing this out at them in their face and they
can say hey no I'm not interested so there's buttons here if we just run this from here you can see oh it's untrained but there's three buttons here now so email phone and I'm okay thanks and I'm okay thanks we can take back to this back to main Loop thing and it's going to ask well what else can I help you with and it takes us back in order to get the entity extraction and lead capture Working correctly we need to come over here we have this sort of same setup We'll be asking for a
valid phone number or an email and then we come over we need to set up an entity so in order to create a new entity I should probably walk you through those steps click here we need to add a new one called email because I'll type email and then we can create this entity here so it's going to capture the email and then down here we can do the same thing where we're going To create a new entity and we're going to call this phone and we're going to make it a phone number then we
can create the entity there and then now we should be good to come back through start this again need to train it I believe yep okay so we're just going to reset it from here run it now let's go email if I go hi my best email is Liam otley at gmail.com then it's going to pass through so it's correctly extracting the email I believe If we pull the sidebar out we can actually see so yes it has extracted the email correctly now we can try it again with the phone if you come up here
and click on the play button start again from the top let me go by the phone route we can go okay I need a valid phone number okay let's go 301 332 and there we go actually let's run this again so the whole point of this entity extraction thing is to be able to not just provide the information that they Want they can actually include it in a larger message so need to do that again so if we go along here phone I can say my best number is 303 82239 okay now it should be
able to extract this phone number and we can check phone there we go it's extracting it correctly now we just need to know what to do with the phone and the email so we're going to be setting up the same sort of system with make as we've done four so what we Can do is go to the templates again back over on make we can go search up webhook start a new web hook yep and then we have the Google Sheets sheet and there we go we have add data to the Google sheet for received
via web hook do the same thing we can start a guided setup in this case I'm just going to add a new web click this is going to be lead YT save this and it's going to give us a link again could copy the address to the Clipboard head back over to voice flow we can click on our get request here we can drop it in this and then you can set up the parameters so I've just done this a little bit at a time we can add the parameter so this is going to be
email this is going to be the email variable that we created this is going to be oh we had another one this one can be phone now in this lead caps you'd probably want to get a first name and a last name or at least just a name to attach to it Very easy step to add in you just put it sort of put it before this section and just make it go first you get the name then you get the email then you send it off to the to the webhook so phone and then
we go phone and then we need to capture we don't need to catch anything we're just looking to send the information off so we have email and phone we have the hook we can paste in there and then we can send the request so let's put in the email liammail.com Send request and then 200 code means it's gone through all good we can hop over to make and then we see we've successfully determined it we can continue I've set up my Google account already as we've done before now actually just to save the sake of
more sheets I can add a new sheet down here and call this lead capture and I can say email phone obviously you want to enrich this more with that first name last name as I just said but also you can take Some of the information within the conversation so you could do a little AI set AI task and you could take all of the messages that have been said during the conversation so far and summarize them down or even just pipe those straight over and send all of the transcript data from the conversation that the
users have into this lead sheet as well so this means you're going to have these a first name last name email phone maybe then you're going to have a Really enriched data set on what they've actually been asking to this bot and based off those questions you can very easily pass it over to the sales team or start off some personalized emails and things like this so it's like really really awesome stuff that you can do at the back of this I've kept the basic within this but I just sort of want to let you
guys know you in the right direction and you guys can go and sort of brainstorm on it I will be doing some More advanced tutorials so if you want to stick around for that I'll be dropping them later probably on the second Channel at some point so be sure to subscribe to the channel and keep an eye out for the second Channel when it comes and so I can pick the spreadsheet here which was incident reporting you guys should make the correct one of course see you didn't reporting and then I can put it on
the second tab of the sheet I can go lead capture I can Continue and then what I need to do is probably send another one over if we go resend come back and then what we can do is click on the Google sheet here we can go we need to redetermine the data structure because I've backed out of it sorry look at searching for it we just pop back over how you know the drill now you can resend that it's going to come back in here and we can continue come over to Google Sheets Google
recording okay okay and then so you may need to Hit the refresh button there as I just did and then we can add the email to that and we can add the phone to that as well so we can continue there I guess the test really is to run a prop query in here and we can resend in order to test it it's working we can send the web hook and then we can click run once we can use the existing data run this and then we should have a pop over great we can just
test it one more time I like to do this a couple of times if you can't Tell we can run this again using existing data and we're going to pop another one and so then to make this lead capture work around the clock I can turn the scheduling on and then we can head back to the scenarios page and we can watch the data come in this one is still on we can head back over here resend and it should pop up there great okay so it's working correctly and now we need to actually test
it with some real data so if I go back here click the Play button I want to do an email I'm going to provide my email my email is Liam gmail.com there we go API called successfully triggered path not connected and boom there we go so it didn't have a phone it's just returned to zero but we did have an email so that's we can get a lead capture set up within your voice flow application sending over to a Google sheet and that I mean the interesting part about this Is that as soon as you've
got the information into your make scenario here you can do so many different things with you've got this big editor you can start to add in all these different flow controls all these different tools the cool thing is that Google Sheets in the build endle once you have this information in make essentially you've got the web of data you can start to manipulate it in all sorts of ways and really connect it to different parts of The business so I'll leave that open to you guys as I said before this is the really cool stuff
once you get a chat bot as a data source and you can start to build stuff off the back of it which is what I really love doing so my final steps here is that if it's a success get out of this editor Escape what we can do is click on this drop it here add action we want to go to a block and then we can go to the back to home so after it's successfully completed we're going to Send it back to home if it fails we can come down here and say failed
and then we can just end here so you can walk this off and then just go to end actions end so that's how you can just end the conversation if something goes wrong but now the key part here is that we've got all of the ads set up we've got the lead capture we've got this main block as well now we just need to do the ad router every time we hit three messages we are going to be sent over to this ad Router so the ad trigger that's just being sent at the moment what
we're going to do is handle a bit of logic here so this is a multi-step process we're going to go if condition up here see if we can sneak this guy in can delete this maybe I'll leave it there for now actually just for debugging sake and we can say if the variable add router is one then we're going to go to block and we can go to add add one and then we can do a no match so this is an Else block so if this is not met then we can add a path
and then I want to go down here we want to do the same sort of thing we just create a logic step if condition actually I might just copy copy this thing paste it in here and delete this this guy on here add router then we're going to go if add router is equal to two then we're going to go to add to even that doesn't work we're going to send it down to here so add relative is Equal to three then we're going to do something here no match we can just send to go
here we can say something like text add browser failed let me just leave them down there then we need to change this to three add router three so if this is seemingly confusing basically as soon as it comes in we have three different ads and we have this toggle which is a variable that we're going to be editing every time we go through an ad so when the Conversation starts we're going to have the ad router set to one as we do here so set add relative to equal to one that means that the first
ad that's going to be shown when we hit three messages which is checked for here is going to be sent up to the ad router then we're going to check what the value of the ad router is if it's one then we're going to go to add one and then if we go to add one we're going to update the add router to two then they're going to show Them the ad we're going to send them back to the main Loop they can continue to chat but we're also setting the messages back to zero so
that when you get up to three messages being sent again it then goes to the add writer but this time the add writer variable is set to two and so it comes through here if the ad router is 31 okay it's not set to one else we're going to go to the no match comes down here we go to the ad writer two and then let's go okay this This checks out we have an ad router value of two then we're going to go to add to and so on then to add two is going
going to set the ad router to three and it's just going to be this continual process and here we have on the lead capture the ad router is set back to one so we're counting the number of messages we are turning to the ad router to check which ad we should be showing and then we're going to be updating the advantage variable so that We're changing the ad that's shown each time now I believe we are in a position where we can actually run this bad boy and check it out in action and wrap this
thing up so if we need to train anything more no hi I'm Nick what would I like to know let's just hi Nick sorry the answers are quicker if I'm not asking them a big fancy question there we go it's giving me a big answer add triggered boom and then we get the first ad so I can claim my thing and it's Going to take me to your man here and then I've clicked this button it was claim your and it's going to take me back oh that should be claim yours now it's taking me
back to the main Loop and it's put me back here and you can see the ad router is now set the two the messages are seen set back to one by this little bit of JavaScript up here and then we can continue to chat to it what are my financing options now the quality of these responses is going to Highly depend on of course your knowledge base so here take this with a grain of salt this is chat GPT generated this isn't some secret source of the industry so keep in mind that this is not
the optimal and this is definitely not what it's going to look like in practice again he's given me a big chat about how the financing options are available then we can ask them again hi Nick let's just keep them going and show some of these other ads hi Nick maybe I'll turn the debugging off so it's not bugging you guys huh add triggered boom take us to the calculator there we go we can capture the lead again and we say hi Nick again just using high neck just to get this conversation a bit quicker so
I'm not sitting here for half an hour a high neck high neck and then add triggered blammo so if we hit okay thanks it's going to take us back to the main Loop but if we click on email I can Go my best email is Liam at liam.com and then bam we've executed this and ideally Liam liam.com we've captured the lead as well and then ideally we can put some kind of column in here which is like transcript and we've also got a whole bunch of data on what the user's been asking about that we
can use for personalized Communications we can even get the transcript and put it through a little AI function and get it too classified into different types so what Kind of customer is this what kind of category within our business and our kind of uh Prospect and database could he fit into okay let's put into this team who's very good with this like a business clients really really cool stuff you can do as soon as you put in the transcript in here and then you can go back to make and you can run AI features on
it and all sorts of things so that's the really cool stuff chatbots may seem a little bit stale but once You're actually providing real value to these businesses by collecting this information and then doing something valuable with it that's what this thing's all about really so that's what we're trying to work you towards but that's been the advanced lead generation bot I hope you've understood how this thing is working you can take this template and swap out the different ads that you're using you can adjust your knowledge base you can change the lead Capture to
be a bit more extensive asking for more information Etc but this is the gist of it now we can share the Prototype if I just pop this over here but we can also as I mentioned before let's get out of the designer here we can publish it I'm going to call this V1 I have the YT lead thing so hi I'm Nick hey Nick what why is solo good for residential and there we go so it seems a little bit better once we've got a wider chatbot Window to look at this in again quality of
the information probably not great because I've used the generator chat CPT to to create the ebook itself but that gives you an idea of how you could see it on a full screen share it with some clients Etc and then what we want to do is embed this widget so I can show you it on a web page if I just pop back over to this take the script tag out save it I can run this up again and now we have it down here here we have Nick hi Nick We can take a look
at some of the ads in here actually let's see what they look like once you're on a web page hi Nick heck Nick hi Nick add trigger claim your free ebook there we go so what you do actually is remove this ad trigger now that we know that it is triggering correctly so you can come back over here Etc pop back to the designer and you can remove this ad trigger in that case you could redeploy Etc one thing I also wanted to point out that's outside of The scope of this video is that you
can take all of the awesome functionality that you build on voice flow like this and you can use their dialog API which I'm showing on screen here you can sort of custom code this a little bit but you can create an amazing lead generation chatbot that the back end really the the functionality of it and how it's supposed to work and using all these AI features and then you can plug that into a custom front-end so if you want to Create a cool lead generation website or something for one of your clients you could create
a a new web page and you could stick this on and create your own custom UI and really bake it into a custom site so you use a voice flows to backend in the logic but then you can create some really really cool front-end address this lead generation chatbot up to feel more than just a little web chatty thing in the corner so I'll leave that with you there I might come Circle Back and cover it in a later video maybe on the second Channel when that's all set up and running but I just thought
I'd drop that in there and there's also a knowledge base API so you can query your knowledge base and not necessarily use it how they've set it up in the no Cody week you can actually call the knowledge base and pull chunks back and have a lot more control over it so that's it for the bot building guys we've really got a lot of rubber on the Road here I hope you guys have hung in there and made a good effort of this or as I said all of the templates are going to be available
on my resource Hub in the description so if you want to clip through to that you can sign up it's free you're going to get access to all of these different templates to download plug and play and use them how you like and with that everyone we've finished the chat by building section of this course we've collected three different Prototyping methods so you have all the tools you need there to spin up something quick and really demo it and we've gone through six different examples of chat to basic customer support and advanced customer support a
basic WhatsApp and staff training chatbot and an advanced one whom we've also done in a basic and an advanced lead generation chatbot so I really think we've covered a lot of ground here and now that you know everything under The sun about how do you build chat Bots the only thing left to cover for you all in this video is how can you actually start to sell these to businesses luckily you've been listening to and learning from the right guy here I do this every single day in my own businesses I run an AI automation
agency which is a business model that I created as I mentioned we've been over a little bit before in the video but I want to give you a little bit more information On it now and exactly what it does and how you can start to take the skills you've created and begin to start selling packages of AI Automation Services Beyond just a little chat bot and going over to really a valuable package of of deliverables for a specific Niche so what is an AI automation agency well the official definition is this the AI automation agency
model is an online business model focused on helping small to medium-sized Businesses to automate their systems and processes with AI technology so a beautiful thing about you making it to the end of this video is that you've just learned basically half of the skill set in terms of delivering services for this what you need to do is combine chatbot building skills with automation platforms like we've been using with make today and zapier and things like that our job is AI automation agency owners is to help businesses to adopt This new technology and really benefit from
the huge gains that they can provide so if that means to apply some of these chatbots to their businesses as new data collection points connect to different parts of their businesses finding repetitive tasks that are being done by their staff members and automating those processes and systems as well the the Bedrock that this business lies on is your ability to connect and draw a direct line as I say On screen how to draw a direct line between the product you create and the money that they're able to save all the revenue they're able to increase
so it's no rocket science there's nothing crazy here you are supposed to identify a part of the business that you can improve or automate and you can draw a direct line from hey this is how many hours someone is going to be spending on this kind of thing per month based on what you've told me I can cut this down to one-fifth Of that and you're going to be saving this much money so that is where you can really start to tap into providing value as an AI automation agency owner and the key to getting
started and really finding success what I found from my own experience and from working with my students I have an original eight incubator program that I'm doing for free at the moment and I've also got thousands 23 000 people in the Discord as well that I'm working with on sort of A more distance basis but the key is that we've found is to focus on one Niche say you become the AI automation expert for real estate agents where you can become the AI automation expert for construction firms and things like this you focus on building
out a package of things that people in this industry you know typically they're going to need and their business could benefit from and then you're able to package this up as sort of a bundle and start to offer this And this might include chat Bots it might include automation systems as we've been going through here as well but the key is to get to that point where you're able to have sort of a nice set of things you can go and be the expert at automating businesses within a particular Niche so if you're interested in
learning more about how to start and build your own AI automation agency you're in the right place because I've got a ton of videos on my Channel about The specific topic book this is really all we're focusing on right now so if you want to be able to take your AI chatbot skills you've got now and convert that over into building a valuable and sustainable business with an AI then this is your ticket as I mentioned earlier in the video I have a two hour full course on how to start an AI automation agency so
if you've got all your chatbot skills from this video you are in the perfect position to head Over to that video and soak all that information up and start building these things and making money so if you want to get any more resources first of all subscribe to the channel if you haven't already and please leave a like down there and mean the world to me I'll put so much work into this video for you guys and if you want to join the biggest community of AI automation agency owners and really bright Sparks and AI
at the moment feel free to join my Discord it's A free community we also have the paid One releasing soon as I've mentioned earlier in the video so if you want to sign up to the free one you can get in there now if you want to join the wait list for the premium one which is releasing soon you can find a link in the description as well if you're watching this video a bit later it's probably going to be available to get into the premium already unless we are full again and then you might
have to Wait a little bit to get in on the next batch but in that case you can also sign up to my resource Hub if you haven't already with all the resources that I mentioned and a ton of other ones on chatbots I've built and things to do with AI automation agencies too and finally you can sign up to my AI automation agency newsletter called over the shoulder where I'm giving you weekly tips and tricks and sort of access to my personal notepad you're looking over the Shoulder on the notes and the things that
I'm picking up each week when I'm running my AI businesses when I'm working with all of my students so I'm giving all of that special sauce directly to my newsletter every week and people are loving it so that is all for the video guys thank you so much for watching and congratulations if you made it this far I think it's a testament to who you are as a person that you've been able to stick with this I'm sure you've Got a notebook full of notes and you're ready to really attack this opportunity because it is
such a huge opportunity for all of us right now who are ready to put in the work so that's all for the video guys thank you so much for watching and I will see you in the next one [Music]
